The recent recrudescence of gonorrhea provides an opportunity to link transmission in neighborhoods and localities to national trends by examining the area under the curve of reported cases for geographically defined regions at each level of data aggregation.
From *Emory University School of Medicine, Atlanta, Georgia
Thoughts and materials for this article were derived in part from discussion at the Consultation on the Control of Neisseria gonorrhoeae Infection in the United States, convened by the Division of STD Prevention, Centers for Disease Control and Prevention, on October 10–11, 2001. The authors thank the conveners and participants for the opportunity to be part of that meeting.
Correspondence: Richard B. Rothenberg, MD, Department of Family and Preventive Medicine, Emory University School of Medicine, 69 Butler St., SE, Atlanta, GA 30333. E-mail: email@example.com
Received for publication November 28, 2001,
revised March 8, 2002, and accepted March 8, 2002.
LONG-TERM EPIDEMIC TRENDS, particularly those with a sustained rise or fall, lend themselves to facile explanations. From a nadir of 193,468 cases of gonorrhea in 1941 (Figure 1), the steady climb during and immediately following World War II was attributed to war time conditions and to the joyous return of servicemen from 1945 to 1947. The subsequent decline in cases from the peak in 1947 to a low point in 1957 was interpreted as a reversion to “normalcy,” as veterans (and others) settled into the welcome calm of the 1950s. A rise in cases began in the early 1960s, with exponential growth beginning in 1965 and plateauing from 1976 to 1980, with over 1 million reported cases annually. This massive increase has often been attributed to the changing social and sexual mores of the 1960s, augmented by the postwar population bulge. Some portion of this increase, however, may be attributable to a major gonorrhea screening effort begun in the early 1970s, though the relationship of that program to the ultimate termination of the increase remains uncertain. 1 The downward trend, first noted in 1981, actually preceded the advent of HIV/AIDS, but the latter's effect on the sexual activity of men who have sex with men is the usual explanation for its continuation. A small uptick in 1984 was short-lived and unexplained and was followed by a sustained decline in reported cases through 1996. This decline has been largely ignored, with the lingering explanations related to gay men's sexual activity. The consistent decrease did not appear to be related to gonorrhea control efforts.
By 1997, the number of reported cases of gonorrhea had returned to the level reported in 1965, although, because the total civilian population is used as a denominator, the rate had decreased from 169.2 to 122 cases per 100,000. A second uptick began in 1998 and, unlike the episode of 1984, appears to have been sustained through 2000 (although there were 899 fewer cases reported in 2000 than in 1999). Whether the explanations for the long-term trends have been right or wrong, they certainly have an air of verisimilitude that comes with sustained and consistent change. Explaining short-term change is more difficult.
Monitoring Gonorrhea Trends
Epidemiologists use several approaches. Surveillance data are often evaluated on a year-to-year basis. Such periodic change is measured by trend line analysis or by the percentage change from the previous year. If, for example, we examine the year-to-year change from 1996 to 2000, we observe a small increase in the reported number of cases, whose trend line, on the scale provided, is barely increasing, but whose slope is significantly different from zero (z = 3.41;P < 0.001) because of the single increase in 1998 (Figure 2 a). 2 That increase, 8.9% from 1997 to 1998, is also displayed by the year-to-year percentage change and makes the remaining changes appear inconsequential (Figure 2 b). Viewed in the statistical perspective of the long-term downward trend that preceded it, this upturn did produce a significant increase of about 12% in the slope of the trend line from 1980 onward (comparing slopes from 1980–1996 and from 1980–2000: z = 2.22, P < 0.05). Thus, the year-to-year approach established a significant change in direction but provided little additional guidance.
An alternative is to examine the cumulative number of cases within a period longer than 1 year, a technique that is a modification of the CuSum plot. 3,4 The area under the curve, calculated as the difference between each year and the baseline year, can be viewed through a movable window. For the period 1996 to 2000, the area under the curve is nonnegligible and sustained (Figure 2 c), yielding an excess of 93,431 cases (or 136,846 cases among those states in which an increase occurred [see below]) over the number reported for 1996. The year 1996 was chosen as the baseline because, as we will point out, changes between 1996 and 1997 presaged those between 1997 and 1998. Viewed in this light, the change evokes a greater need for response, though the situation is admittedly not as blatant as for the period 1975 to 1980, when 4.2 million cases were generated in excess of those occurring a decade earlier.
The national data provide some opportunity for assembling more specific evidence of change. Comparing each year to 1996, we note an excess of states that reported an increase (Table 1); this excess occurred during the year 1997 as well, when there was in fact a decline in total reported cases. For all four comparisons, the number of states with increases was greater than the number with decreases, and this phenomenon was significant by the sign rank test, with one exception (the 1999–1996 comparison). Since only one of the four comparisons was significant with the binomial approach, statistical testing does not fully confirm a departure from random fluctuation, but the consistency of the finding over the time period is of importance in judging its credibility.
The Loci of Change, 1996 to 2000
To determine the major contributors to these increases, the states were rank-ordered for excess cases by subtracting the number of cases reported in 1996 from the number of cases in each of the other years. This difference was used to sort the cases from largest (positive) to smallest (negative) change. The rank order for the 1998-to-1996 comparison was used as the standard so that the rank of each state in the other years could be compared to its rank during the year of greatest increase (Table 2, upper portion). The median rank for each state for the four comparisons was then computed. The data reveal that, although there were fewer cases in 1997 than in 1996, the increase in four of the top five states in 1998 and thereafter was presaged in 1997 (Ohio is the exception). Furthermore, the top seven states plus number 11 (Pennsylvania) show consistent increases that are maintained over the 4 years of observation. Four of the states (Missouri, North Carolina, California, and Georgia) were more volatile. Texas was the undeniable leader, ranking first for each comparison and having the largest cumulative contribution of excess cases (it accounted for 24% [32,778/136,346] of the total increase observed among states that increased during the study period). Notable by their absence are such states as New Jersey and Alabama and the District of Columbia. New York, which was ranked third for the 1997-to-1996 comparison, had substantial decreases thereafter. The state of Florida, though not included by the decision rules used, warrants special mention after perusal of the data (only a portion of which are shown [Table 2]), since its contribution entered later than those of the other states.
Using the same process for the selected cities that are reported nationally (Table 2, lower portion), the top 12 cities provide placement, at least within some of the states, of the areas of excess. Though several of the city rankings are ambiguous (Chicago, Houston), and one city (Phoenix) does not have a corresponding state in the top 12 (Arizona was 14th), the general process of detecting focal contributions by states and focal contributions of cities within states is evident.
Such shifts in reported cases can be readily monitored at the national level and made available to states and localities, but it is at this point that the national data give out, and we arrive at a disconnection between smaller-area reporting (that is, the places where gonorrhea is actually transmitted) and the national system. To make such a connection, we briefly review the current hypotheses regarding the dynamics of STD transmission.
STD Transmission Dynamics Revisited
Current theory posits the existence of small groups (ambiguously referred to as “core groups”5) that are responsible for most transmission, both inside and outside their boundaries. The existence of such groups was first proposed by Yorke and Hethcote 6 in relation to gonorrhea, and empirical studies have provided verification of such groups in a number of settings and for a number of conditions. 7–15 Though the definition of such groups has been fuzzied by alternative usages, for purposes of this discussion core groups may be thought of as a heterogeneous network whose primary markers of cohesion are geographic contiguity (for example, neighborhoods) or social proximity.
Evidence has recently accumulated that the size and network configuration of core groups are related to endemic and epidemic phases 16 of STD transmission. 14,17–26 Differing network characteristics thus provide a direct mechanism for placing core groups along the spectrum of transmission potential. The “summing” of such core groups provides a direct mechanism for assessing the meaning of secular trends in reported gonorrhea morbidity.
Completing the Connection
To provide the links between national and local disease reporting, states would need to carry on the process by examining the area under the curve (and its potential causes, including possible artifact) for their larger jurisdictions (counties, cities), and within these, their smaller geographic designations (census tracts, zip codes). If geocoding of addresses is available, the process may be continued to the neighborhood level. It is likely that concentrations of cases at each level contribute to the prominence of the level above it. If the underlying theory is correct, such concentrations signal the presence of core groups “on the ground.” When the surveillance analysis is brought down to the neighborhood level, the link between national and neighborhood data becomes manifest. Finally, states and localities are obliged to “explain” the change (artifact, heightened endemic transmission, outbreak) and to act on their assessment.
Collation of local data can thus provide meaning for the national trend. National totals alone, viewed as a summation of sex acts, provide little insight. Viewed as a summation of groups that transmit disease, 27 national data reflect the spectrum of current transmission patterns. Overall change in national trends is a weighted average of changes within groups, where the weights are proportional to the number of groups at each position on the spectrum. A particular city may account for its own increase by attributing it to, say, a change in risk-taking by men who have sex with men, but the national increase results from many changes in many areas, some of which may share that etiology, and some of which may have other causes.
A national strategy follows directly from such linkage: observe the current upswing and verify its reality (Table 1); identify the harbingers of change in the period before true increase began; find the areas of consistent increase (Table 2); and focus on the states with important changes, the cities within them, and the smaller areas within the cities. Direct interventions, such as network-informed contact investigation, have the potential to interrupt transmission within these areas. 14 The targeted deployment of such interventions has the potential for rapid impact on the national trend.
Here is the fanciful part: all that is required is to redirect resources, personnel, and energy to the areas of consistent increase (or even to areas with high levels of entrenched endemicity). Those familiar with the bureaucratic structure of STD control will see at once the substantial obstacles to such a recommendation. States and locales, currently underfinanced, understaffed, and undertrained, would express justifiable concern over a shifting base of support. The fact that the target is moving (for example: Florida, upward; New York, downward) is a further complication. The ultimate priority that the control of gonorrhea might be given is yet another. We are thus at some distance from implementing such an approach, whose current empirical verification rests on a small sample of programmatic data. Perhaps the first step in providing the missing links between federal and local data is an intensification of epidemiologic and computer training for states and localities, with the goal of meaningful analysis of local events.
1. Henderson RH. Control of sexually transmitted diseases in the United States: a federal perspective. Br J Vener Dis 1977; 53: 211–215.
2. Division of STD Prevention. Centers for Disease Control and Prevention. 2000 Sexually Transmitted Disease Surveillance Report. 2001. Available at: http://www.cdc.gov
3. O'Brien SJ, Christie P. Do CuSums have a role in routine communicable disease surveillance? Public Health 1997; 111: 255–258.
4. Wohl H. The CuSum plot: its utility in the analysis of clinical data. N Engl J Med 1977; 296: 1044–1045.
5. Thomas JC, Tucker MJ. The development and use of the concept of a sexually transmitted disease core. J Infect Dis 1996; 174 (suppl 2): S134–S143.
6. Yorke JA, Hethcote WHJ, Nold A. Dynamics and control of the transmission of gonorrhea. Sex Transm Dis 1978; 5: 51–56.
7. Zenilman JM, Bonner M, Sharp KL, Rabb JA, Alexander ER. Penicillinase-producing Neisseria gonorrhoeae
in Dade County, Florida: Evidence of core group transmitters and the impact of illicit antibiotics. Sex Transm Dis 1988; 15: 45–50.
8. Zenilman JM, Ellish NJ, Fresia A, Glass G. The geography of sexual partnerships in Baltimore: applications of core theory dynamics using a geographic information system. Sex Transm Dis 1999; 26: 75–81.
9. Alvarez-Dardet C, Marques S, Perea EJ. Urban cluster of sexually transmitted diseases in the city of Seville, Spain. Sex Transm Dis 1985; 12: 166–168.
10. Arya OP, Rees E, Turner GC, et al. Epidemiology of penicillinase-producing Neisseria gonorrhoeae
in Liverpool from 1977 to 1982. J Infect 1984; 8: 70–83.
11. Potterat JJ, Rothenberg RB, Woodhouse DE, et al. Gonorrhea as a social disease. Sex Transm Dis 1985; 12: 25–32.
12. Rothenberg RB. The geography of gonorrhea: empirical demonstration of core group transmission. Am J Epidemiol 1983; 117: 688–694.
13. Rothenberg RB, Potterat JJ. Temporal and social aspects of gonorrhea transmission: the force of infectivity. Sex Transm Dis 1988; 15: 88–95.
14. Rothenberg RB, Kimbrough L, Lewis-Hardy R, et al. Social network methods for endemic foci of syphilis: a pilot project. Sex Transm Dis 2000; 27: 12–18.
15. Rosenberg D, Moseley K, Kahn R, et al. Networks of persons with syphilis and at risk for syphilis in Louisiana: evidence of core transmitters. Sex Transm Dis 1999; 26: 108–114.
16. Wasserheit JN, Aral SO. The dynamic topology of sexually transmitted disease epidemics: implications for prevention strategies. J Infect Dis 1996; 174 (suppl 2): S201–S213.
17. Friedman SR, Neaigus A, Jose B, et al. Sociometric risk networks and risk for HIV infection. Am J Public Health 1997; 87: 1289–1296.
18. Kretzschmar M, Morris M. Measures of concurrency in networks and the spread of infectious disease. Math Biosci 1996; 133: 165–195.
19. Morris M, Kretzschmar M. Concurrent partnerships and the spread of HIV. AIDS 1997; 11: 641–648.
20. Potterat J, Zimmerman HP, Muth SQ, et al. Chlamydial transmission: concurrency, reproduction number and the epidemic trajectory. Am J Epidemiol 1999; 115: 1331–1339.
21. Potterat JJ, Rothenberg RB, Muth SQ. Network structural dynamics and infectious disease propagation. Int J STD AIDS 1999; 10: 182–185.
22. Potterat JJ, Phillips-Plummer L, Muth SQ, et al. Risk network structure in the early epidemic phase of HIV transmission in Colorado Springs. Sex Transm Infect 2002; 78 (Suppl 1): i159–163.
23. Potterat JJ, Muth SQ, Rothenberg RB, et al. Sexual network structure as an indicator of epidemic phase. Sex Transm Infect 2002; 78 (Suppl 1): i152–158.
24. Rothenberg RB, Sterk C, Toomey KE, et al. Using social network and ethnographic tools to evaluate syphilis transmission. Sex Transm Dis 1998; 25: 154–160.
25. Rothenberg RB, Long D, Sterk C, et al. The Atlanta urban networks study: a blueprint for endemic transmission. AIDS 2000; 14: 2191–2200.
26. Rothenberg RB, Baldwin JA, Trotter R, Muth SQ. The risk environment for HIV transmission: results from the Atlanta and Flagstaff network studies. J Urban Health 2001; 78: 419–432.
27. Holmberg SD. The estimated prevalence and incidence of HIV in 96 large US metropolitan areas. Am J Public Health 1996; 86: 642–654.