THE CONCEPT OF A CORE GROUP is central to sexually transmitted infection (STI) epidemiology. 1 A core group is one in which each infectious person infects, on the average, one or more susceptible people during a given period. By definition, core groups maintain the organism’s viability in the host population. Should the reproduction number, which represents the average number of newly infected contacts generated by infectious persons, drop below one, the organism, incapable of reproducing sufficiently, will die out. 2
The idea that STI transmission is perpetuated by a relatively small group of individuals directly through their own sexual activities and indirectly through those of their sex partners is attractive for both research and prevention. The mathematical origins of core group theory facilitate very clear descriptions of research questions that need to be answered if “the interplay between the variables that determine the course of infection within an individual and the variables that control the pattern of infection within communities of people” is to be understood. 2,9 In terms of prevention, successful interventions in the core group will lead to larger reductions than expected in STI incidence relative to the number of people in the core groups. In fact, this effect was observed after the introduction of a gonorrhea screening program in the United States, during which there was a 10% increase in detected cases. In the following years, the incidence of gonorrhea decreased by 20%. 3
The need for more refined, targeted prevention programs is greater now that it ever has been. A chlamydia control program introduced in Manitoba in 1988 included legislated reporting of cases to public health officials, development of management and treatment guidelines, selective screening, and partner notification for individuals with positive test results. Rates of chlamydia in women decreased substantially from 778 per 100,000 in 1988 to 266 per 100,000 in 1996, after which the incidence rose slightly to 351 per 100,000 in 1999. This trend is similar to that for the whole of Canada, and in 1999 the rate was 123.8 cases of chlamydia per 100,000 women (STD Prevention and Control, Health Canada, in press), 4 exceeding the national goal of fewer than 80 per 100,000 for the year 2000. 5
The initial decreases in chlamydia and the subsequent rises in cost per case detected have been found in other jurisdictions where screening has been implemented. 6–8 In Manitoba, the proportion of positive chlamydia tests dropped substantially from 8.3% in 1988 to 3.2% in 1999. In Quebec, the percentage also dropped from 4.3% in 1991 to 2.8% in 1998. 9 (Personal communication, Dr. Laurent Delorme, November 1999.)
Infections detected during the screening process in individuals motivated to present for annual physical examinations are believed to represent the cases most susceptible to elimination. Reservoirs of infection remain concentrated in STI core groups. Social, cultural, and economic marginalization of these populations also may result in less contact with the public health system. Investigation of new prevention methods in these groups is essential as the number of cases decreases and screening becomes less cost effective.
The fact that sexual networks play a significant role in core group dynamics was recognized as early as 1981, 10 when people connected directly and indirectly through sexual intercourse were grouped in lots, six of which accounted for 20% of all the STI cases in Colorado Springs. Since then, this approach has been advocated as an important method for addition to the available tools used to describe STI transmission. 1,11–13 It is logical to infer that large networks contain core groups because there is evidence of infection transmitted from an index patient to his or her sex partners, who in turn transmit the infection to one or more people, who in turn are linked with yet more secondary cases, forming a large network. Interruption in transmission will fragment the large networks within which STI proliferates and from which STI are disseminated. This strategy will be more effective and efficient in reducing rates of STI than intervening in smaller sexual networks of only two or three people or applying mass prevention strategies.
Despite the potential of social network analysis to add much to the current understanding of STI transmission, 14,15 this area of research is still in its infancy. Not only are more network data needed, 16 but answers to basic methodological questions of sampling populations and evaluating bias also are required. 17 The objective of the current research, one of several studies in a larger project, 18 was to evaluate four methods of easily identifying large sexual networks containing individuals with chlamydial infection. These results also could be used in public health practice to prioritize partner notification of the clients likely to be members of large networks.
The authors selected four subgroups of STI cases that they believed would reveal the largest networks. The rationale for selecting these groups originated from an earlier study conducted to determine the feasibility of constructing sexual networks of core group members from routinely gathered contact tracing data. 19
The first group consisted of individuals who had two or more positive chlamydia test results in the 6-month study period (repeaters). Individuals who had laboratory-confirmed coinfection with gonorrhea and chlamydia formed the second group. The third group comprised both coinfected individuals and those with repeated chlamydial infections. The fourth group consisted of individuals selected by random sampling of all specimens that had positive test results. The number and size of the networks constructed by these four different methods then were compared against those of networks constructed by use of information for all cases and contacts reported during the study period.
Testing for genital Chlamydia trachomatis infections in Manitoba is centralized at Cadham Provincial Laboratory (CPL), which performs 95% of all chlamydia diagnostic and screening tests (approximately 60,000 tests annually). Chlamydiazyme (Abbott Laboratories, Mississauga, Ontario, Canada) and Gonorrhea Pace 2 (GenProbe, San Diego, CA), nonamplified, were the diagnostic tests in use at the time of this research. Selective screening was recommended for all women with multiple sex partners or a recent history of or exposure to an STI, and for women undergoing therapeutic abortions or intrauterine device insertion. Screening of all women younger than 25 years was added in 1989 to the Canadian guidelines for chlamydia screening. 20 Screening was not recommended for women presenting for annual or prenatal examination who did not fit any of the aforementioned categories. 21 Computerized records for each specimen contained information on the types of tests performed, the results, the attending physician, the patient, including the personal health insurance number, and the specimen dates.
In Manitoba, physicians and nurses practicing in medical clinics, hospitals, and health centers on Indian reserves screen, diagnose, and treat patients with STI. Once a positive laboratory report is received at Manitoba Health Communicable Disease Control (MCDC), the responsible public health unit is informed, and a public health nurse contacts the physician. The physician completes the sexually transmitted disease notification form, which is entered into a computerized database by staff at MCDC. If the physician wishes, he or she may interview the patient to obtain information on his or her sex partners. Sexual partners of a patient are requested for a standard time frame since the onset of symptoms or, if the patient is asymptomatic, for the 3 months before the positive laboratory test. If the physician wishes, the interview is conducted by a public health nurse, who fills out the sexual contact information on the notification form. The form is forwarded to MCDC, and the information entered on the computer along with the laboratory and demographic data.
Partner notification is completed in one of three ways. First, the contacts reported by the index patient can be notified of their exposure by local public health nurses or by the primary caregiver. Second, the index patient, if judged competent, can inform his or her own partners of their exposure. The healthcare provider may still note the names to ensure that the partners present for testing and treatment. If the partners fail to present, the provider locates and notifies the partners herself. Third, if the client is assumed to be capable of locating and notifying his or her own partners, it is possible that the provider may not have reported the names of the contacts to MCDC. In this case, MCDC would not have a record of the contact’s name.
Partner notification for women with chlamydia detected by screening may not have been as complete as for those detected by diagnostic testing because infection may have taken place months previously, complicating the location of previous sex partners. In the current study, compliance with notification system was good, demonstrated by the fact that contact follow-up evaluation did not reveal previously tested, positive individuals unknown to public health officials.
Computerized registries of all laboratory-confirmed patients and their named contacts are maintained by MCDC. The computerized databases were designed to fill the dual purposes of recording epidemiologic data on gonorrhea and chlamydia cases and providing a reference source for client and contact care. They contain one record for each notification of a disease episode to Manitoba Health. Each record contains detailed demographic, laboratory, contact, treatment, symptom, and healthcare provider information for each patient. The contact database contains the name, birth date, age, gender, and address; whether the contact was located, tested, and treated; and the disease to which the contact was exposed.
Data from the CPL computerized record system on each specimen from November of 1997 to May of 1998 were downloaded electronically to a personal computer. The ASCII files were imported into Epi Info (v. 6.2, 1996, Centers for Disease Control, Atlanta, GA), and the dates were recoded to allow for calculations using dates. The records of all individuals with positive results for chlamydia only were selected regardless whether they had an additional positive or negative test result for gonorrhea. Four lists of specimens were created that were not mutually exclusive because the aim was to assess simple, convenient markers of large networks rapidly. The first list included repeaters who had more than one positive test for chlamydia, and none for gonorrhea, during the study period. Time periods between positive tests were not specified, for the reason stated earlier. The second list included all individuals with a single specimen simultaneously positive for both gonorrhea and chlamydia at any time in the study period. All repeaters and coinfected individuals composed the third list. No one had more than one specimen positive for both chlamydia and gonorrhea. The fourth list comprised randomly selected chlamydia-positive individuals generated using Rsample, a program written in Epi Info, which retrieved a 10% sample of all records.
The complete sexual network was constructed by linking computerized specimen records from the laboratory database with the individual electronic records at MCDC using the personal health insurance number (PHIN). A PHIN is assigned to each resident of Manitoba for obtaining medical services under the publicly funded health insurance program. Continuous validation of the computerized information occurred because each individual’s demographic information was checked against the Manitoba Health Patient Registry file each time a laboratory sample was received. The patient registry contains the demographic information for all the residents of Manitoba who are eligible for publicly funded healthcare services. Because the health insurance system does not require the payment of premiums and provides universal health insurance, including all basic inpatient and outpatient services, nearly all the residents of Manitoba are registered.
The laboratory records were matched with the notifiable disease records from August 1, 1997 through May 1998. Inclusion of the data from the preceding 3 months allowed capture of data on the index patients who had named individuals testing positive after November of 1997. These index patients, or “infectors,” are more important in transmitting infections than those they infected. 22 The identification number assigned to each registered case with an STI was used to link index patients to their contact(s) in the contact database. The researchers verified whether the named contacts had tested positive and now were also registered as cases. This was accomplished by cross-matching the names in the database of laboratory-confirmed cases using the following criteria in each of two circumstances. First, the first name and surname or alias matched exactly and (1) two of three items (day, month, year) of the birth date were identical, (2) the month and day were identical but reversed, or (3) the addresses (including the house or apartment numbers) were identical. Second, if the names did not match exactly, as with shortened first names or misspelled last names (e.g., “Michael” shortened to “Mike,” or “Taylor” spelled “Tailor”), then one of the three preceding conditions had to apply or the ages of the two records matched (allowing for the year in which the events were reported) and the address matched exactly. This algorithm, validated in an earlier feasibility study, 19 resulted in a complete list of individuals.
Unique identifiers then were assigned to each individual and organized into a linked list format (pairs of links). Initial attempts at network construction with UCINET IV (©1996 Analytic Technologies, Harvard, MA) failed because of the large data set. An alternative program for large data sets (PAJEK [Program for large network analysis], ©1996 Vladimir Batagelj, Andrej Mrvar, http://vlado.fmf.uni-lj.si/pub/networks/pajek/) was in development at the time of this study. To test the accuracy of PAJEK, a smaller data set was imported into both UCINET IV and PAJEK. Because both programs produced identical results, the complete data set was imported into PAJEK for analysis.
In networks containing individuals from the random sample linked directly or indirectly by sexual intercourse, the repeater and coinfected groups were identified. The network size for each group and for the whole data set were compared using Kruskal-Wallis tests for nonparametric data to assess differences between two groups.
From November 1997 until May 1998, the CPL mainframe database contained 34,312 unique test results for chlamydia and gonorrhea tests. Of these, 1351 were records containing positive test results. Individuals who had tested positive were matched with those in the notifiable disease data set, which generated a database of all 4544 confirmed cases and contacts. Of these, 979 were either index patients with positive test results before November 1997, but whose named partners subsequently had positive test results during the 7-month study period, or contacts named by people at the very end of the study period for whom no demographic data were available because they were outside the defined study period. The remaining 3565 people comprised cases (n = 1141) and contacts (n = 2424). The number of individuals did not match the number of tests because of repeat testing, and because of lag times in data entry into the notifiable disease database the week after the study period cutoff. Of the cases and contacts for whom information was available, 1565 (43.9%) were women, and 2,000 (56.1%) were men.
Analysis identified 1705 networks containing 1 to 82 individuals (Table 1). The four comparison groups consisted of 93 repeaters, 76 individuals coinfected with gonorrhea and/or chlamydia, 144 repeaters and coinfected individuals combined, and 120 randomly selected individuals. These were drawn from the 1351 patients tested at CPL who had positive chlamydia test results. Repeaters represented 6.9% of all chlamydia-positive individuals, of whom 82 had two positive tests, 8 had three positive tests, and 3 had four positive tests. Coinfected individuals represented 5.1% of all chlamydia-positive individuals. Coinfected individuals and repeaters grouped together represented 9.1% of all chlamydia-positive cases. Eight individuals had both coinfection with gonorrhea and a second positive test for chlamydia during the study period. Five of the randomly selected individuals belonged to the coinfected group, whereas 10 individuals were repeaters, and 1 individual belonged to all four groups (Figure 1 D). Individuals in the random sample belonged to 114 of the 1705 networks. The repeaters and coinfected cases were members of 76 and 57 networks, respectively, and the combined group of coinfected and repeater individuals were members of 123 networks. The network sizes yielded by the four groups are shown in Table 1.
Proportionately, individuals from all four groups belonged to a significantly higher number of the larger networks than contained in the total population (Table 2). The mean and median network sizes increased in order for the random sample, repeaters, and coinfected individuals, with the combined repeater and coinfected groups being smaller than the coinfected group alone. The number of networks represented decreased from 6.7% for the random sample to 4.5% for the repeaters to 3.3% for the coinfected cases, but was much higher for the combined coinfected and repeater individuals. The proportion of large networks consisting of 10 ten or more people (n = 23) represented in the samples rose from 26%, yielded by the random sample, to 47.8%, yielded by the coinfected individuals and repeaters.
This research was designed to evaluate methods for identifying large sexual networks. Although the coinfected individuals represented the smallest comparison group, they were members of the largest networks. They also were members of the smallest number of networks, which supports the core group concept that very few individuals have a disproportionately large effect on STI transmission. However, the repeaters and coinfected individuals together formed the group through which the highest proportion of large networks was identified. Routine partner notification of people with repeated infections and of those coinfected with gonorrhea and chlamydia afforded access to nearly half of the identifiable large networks. Interrupting the transmission of STIs in large networks with long chains of successful transmission will avert larger numbers of infections than intervening in smaller networks. The advantage of targeting coinfected individuals and repeaters is that these characteristics are more easily available from laboratory test results and simple STI notification systems and do not require detailed demographic information, which seldom is available to public health professionals before the patient is interviewed. The authors suggest that giving repeaters and coinfected people higher priority for STI education, partner notification, and management will further reduce STI incidence, particularly in areas where the incidence has leveled off.
The ability to distinguish between individuals with large networks is essential also for research purposes. The need to study large networks that may contain STI core groups gives rise to a multitude of questions concerning interactions among the hosts, the organisms, and the sociosexual environments, all of which shape STI epidemics. To answer questions on how and why sex partners are recruited into a network, the effects of time, the relation between sexual networks and social networks, and the effects on the organism of passing through different host immune systems, networks of substantial size are required for results to be meaningful. The use of coinfected people and repeaters as “seeds” from which to sample for primary research studies should be particularly useful, allowing the researcher to select more respondents with larger networks rather than use random sampling methods.
There are a number of reasons why all four methods yielded a higher proportion of larger networks than contained in the total population. Logically, the random sample of individuals yielded a higher proportion of larger networks than were present in the whole population. 17 The individuals themselves had an equal chance of being selected, but selection was not stratified by the networks to which they belonged. Therefore, the chances of selecting a person from a large network at random are greater than selecting a person from a smaller one.
Because the repeaters had repeated positive test results over the study period, there were multiple opportunities to collect information on sex partners at each positive test, which resulted in the higher numbers of contacts for repeaters than for nonrepeaters. Although the repeaters may have named the same contact repeatedly, information on these contacts likely was more complete because different information usually was gathered at different interviews. This in turn would have led to a higher success rate in partner identification, location, and testing. The contacts, if positive, would lead again in turn to yet more contacts of their own and thus to larger networks. Also, the nurses may have had additional incentive to locate and treat sex partners of repeaters to prevent reinfections. It also should be noted that people infected repeatedly with chlamydia may be more likely to experience and recognize their symptoms and present for care, and they may be more motivated to give more complete partner information to public health nurses, in attempts to protect themselves from another infection.
Common practice demonstrates that individuals with both gonorrhea and chlamydia coinfection are a high priority for partner notification because they are infected with two notifiable diseases, not only one. This may enhance nurses’ effort and thoroughness in all aspects of the partner notification process, resulting in a higher number of contacts who test positive. Additionally, people who have gonorrhea are more likely to have their partners notified than those with chlamydia. 23 Finally, people infected with both organisms are far more likely to report symptoms, 19 and therefore to present, either as a result of notification or on their own, to healthcare providers.
All of the preceding factors affect the quality of the larger networks under study. For example, it is possible that if more complete information were gathered on patients with dual infections, the analysis would reveal only networks of people who are adept at transmitting both infections, and who may be involved in commercial sex. Therefore, it would be impossible to suppose that networks of people with chlamydia who do not have gonorrhea would be of similar size or structure. The networks in this research were identified by the partner notification techniques used in Manitoba. Different practices and techniques used in other jurisdictions may not produce comparable results, even in similar STI populations. 24 Most important, contact-tracing data and questionnaire research will always show a partial network because clients with STI may not know, remember, or divulge complete contact information. 25
The extent of the bias introduced by incomplete data in this research is not quantifiable. It is possible that the contacts named represent the people with whom the index patient engaged in sex most often, or the contact may have been a person regarded as a “steady” partner. It is equally likely that those not named were anonymous sexual partners who could represent the highest risk in terms of disease transmission. Also possible is that the unnamed contacts are those most feared by the index patient, whether because they provide the index case with illegal drugs or because they would become violent if they suspected that the index patient had provided their names to public health authorities. For these reasons, the authors think it inadvisable to assume that all sexual partners have an equal opportunity of being named. Nevertheless, unnamed individuals may be the most important STI transmitters 24 because they usually are not contacted by public health authorities and may remain infectious for a long time.
Despite the aforementioned limitations, these data collected by experienced public health nurses provide the most complete 26 and consistently gathered data available on sexual networks in Manitoba over time. In addition, the unique combination of a central computerized public health laboratory, a central notifiable disease registry including both cases and contacts, and a publicly funded healthcare system that allows the validation of records with a single population health insurance file are invaluable components in sound research and prevention of STI in sexual networks.
These data indicate that members of the larger networks clearly promote STI transmission and may constitute “core groups.” Most of the people in the interior of the large networks had more than one sex partner, at least one of whom had documented chlamydial infection during the study period (Figure 1). In fact, it is only in this way that large networks can be generated because sex partner information can be collected only from laboratory-proven cases. At least one of a patient’s sex partners must test positive for the partner notification process to continue, and for the network to grow. It is important to note that individuals at the periphery of the networks had sex partners with no laboratory-confirmed infection because their infection may have occurred subsequently to the study period or, rarely, two infected patients may have named the same partner, who may never have undergone testing. Although laboratory-confirmed infection in both the index patient and in at least one named partner indicates membership in a core group, specific proof that the organism is transmitted to the partner from the index patient is problematic. Despite the fact that future analysis of genotyping results will partially address this transmission, prior infection of a sex partner by another index patient may have taken place. Although this complicates the precise identification of the transmitter, it reinforces the theory that the larger sexual networks are core groups. Yorke et al 3 first defined a core in 1978 as being a group in which a “significant” number of preemptive infections take place when infectious individuals contact people who are infected already by different sources. These authors also stated that preemptive infections are substantial when the prevalence of infection in the group exceeds 20%. All the networks of 10 or more people have been shown to exceed this prevalence rate, most by a substantial margin. 18
Ghani et al 17 found that when using a simulated population of people within networks, contact tracing methods identified the large networks in the population, but underestimated the network dimension (size) when the number of contacts traced was low (about 50%). From a purely methodological point of view, this finding is reassuring. Size is more accurately estimated when the number of contacts traced rises above 50%. This finding is consistent with the current data, in which larger networks were identified when repeaters and people coinfected with both gonorrhea and chlamydia may have been interviewed more thoroughly for their partners, as compared with the random sample of individuals who had only one episode of chlamydia. Additionally, the results of Ghani et al 17 suggest that the network sizes found in the current study are underestimates of the networks’ real size.
The primary goal of this research was to evaluate methods of accessing large sexual networks so the dynamics of transmission and the genetic types of sexually transmitted organisms within networks could be described. Selecting from a laboratory database individuals who had coinfections with both gonorrhea and chlamydia and those who were repeaters proved to be the most effective technique, resulting in the highest number of large networks for the lowest number of individuals traced. Although public health practice and the clinical features of a particular STI affect network dimension and structure, the sampling described in this discussion may be a valuable tool for investigating sexual networks.
1. Brunham RC. Core group theory: a central concept in STD epidemiology. Venereology 1997; 10: 34–39.
2. Anderson RM, May RM. Infectious Diseases of Humans: Dynamics and Control. 1st ed. Oxford, England: Oxford University Press, 1992.
3. Yorke JA, Hethcote HW, Nold A. Dynamics and control of the transmission of gonorrhea. Sex Transm Dis 1978; 5: 51–56.
4. Division of STD Prevention and Control. 1998/1999 Canadian Sexually Transmitted Diseases (STD) Report. Can Commun Dis Rep 2000; 26 (suppl 6): 2–5.
5. Anonymous. National goals for the prevention and control of sexually transmitted diseases in Canada. Can Commun Dis Rep 1997; 23S6: 2-1–2-21.
6. Herrmann B, Egger M. Genital Chlamydia trachomatis
infections in Uppsala County, Sweden, 1985–1993: declining rates for how much longer? Sex Transm Dis 1995; 22: 253–260.
7. Sellors JW. Screening for chlamydial infection: taking stock. Can Fam Physician 1995; 41:188–92, 196–9.
8. Centers for Disease Control and Prevention, Division of STD Prevention and Control. Sexually Transmitted Disease Surveillance 1997. Atlanta, GA: Centers for Disease Control, 1998.
9. Delorme L. Dépistage et diagnostic des MTS: êtes-vous efficace? Le Médicin du Québec 1999; 34: 47–51.
10. Potterat JJ, Rothenberg RB, Woodhouse DE, Muth JB, Pratts CI, Fogle JS. Gonorrhea as a social disease. Sex Transm Dis 1985; 12: 25–32.
11. Rothenberg RB. Model trains of thought. Sex Transm Dis 1997; 24: 201–203.
12. Kretzschmar M, van Duynhoven YT, Severijnen AJ. Modeling prevention strategies for gonorrhea and chlamydia using stochastic network simulations. Am J Epidemiol 1996; 144: 306–317.
13. Laumann EO, Youm Y. Racial/ethnic group differences in the prevalence of sexually transmitted diseases in the United States: a network explanation. Sex Transm Dis 1999; 26: 250–261.
14. Granath F, Giesecke J, Scalia Tomba G, Ramstedt K, Forssman L. Estimation of a preference matrix for women’s choice of male sexual partner according to rate of partner change, using partner notification data. Math Biosci 1991; 107: 341–348.
15. Rothenberg RB, Potterat JJ, Woodhouse DE. Personal risk taking and the spread of disease: beyond core groups. J Infect Dis 1996; 174 (suppl 2): S144–S149.
16. Klovdahl AS. Social networks and the spread of infectious diseases: the AIDS example. Soc Sci Med 1985; 21: 1203–1216.
17. Ghani AC, Donnelly CA, Garnett GP. Sampling biases and missing data in explorations of sexual partner networks for the spread of sexually transmitted diseases. Stat Med 1998; 17: 2079–2097.
18. Wylie JL, Jolly AM. Patterns of chlamydia and gonorrhea infection in sexual networks in Manitoba, Canada. Sex Transm Dis 2000; 28:1:14–24.
19. Jolly AM. Sexually Transmitted Disease Core Group Membership in Manitoba: Strategies for Definition and Description of Risk Markers. PhD dissertation. Winnipeg, Manitoba: University of Manitoba, 1998.
20. Anonymous. Canadian guidelines for screening for Chlamydia trachomatis
infection. Can Dis Weekly Rep 1989; 15 (suppl 5): 1–13.
21. Communicable Disease Control. Sexually Transmitted Disease Guidelines. Winnipeg, Manitoba: Manitoba Health, 1987.
22. Hethcote HW, Yorke JA, Nold A. Gonorrhea modeling: a comparison of control methods. Math Biosci 1982; 58: 93–109.
23. Rasooly I, Millson ME, Frank JW, et al. A survey of public health partner notification for sexually transmitted diseases in Canada. Can J Public Health 1994; 85 (suppl 1:) S48–S52.
24. Day S, Ward H, Ghani A, et al. Sexual histories, partnerships, and networks associated with the transmission of gonorrhea. Int J STD AIDS 1998; 9: 666–671.
25. Brewer DD, Garrett SB, Kulasingam S. Forgetting as a cause of incomplete reporting of sexual and drug injection partners. Sex Transm Dis 1999; 26: 166–176.
26. Alary M, Joly JR, Poulin C. Gonorrhea and chlamydial infection: comparison of contact tracing performed by physicians or by a specialized service. Can J Public Health 1991; 82: 132–134.