A LANDMARK REPORT from the Institute of Medicine (IOM) on sexually transmitted diseases (STDs) estimated the 1994 cost of STDs and their immediate sequelae to be $10 billion. 1 While many adverse health consequences of STDs have been identified, 2–13 near-term healthcare utilization subsequent to an STD diagnosis has not been characterized or quantified for any broad general population of women. Such information is crucial to estimating the societal and institutional burden of the problem and assessing the feasibility of studying utilization of health resources as an outcome in trials of STD prevention interventions directed at such populations.
The present investigation was conducted to inform planning of a randomized control trial of a specific educational intervention, Practicing Safer Sex Today (Psst!), which involves adult female Kaiser Permanente Medical Care Program (KPMCP) members who have had an STD diagnosed. The primary outcome in the Psst! trial will be consistency and appropriateness of condom use. However, healthcare utilization subsequent to the STD diagnosis will be a secondary outcome, as will rates of documented STD infection and the emotional burden of these infections. Planning such investigations has been hampered by a lack of information on the pattern of resource utilization that might be expected in patients with a diagnosed STD who receive usual care (i.e., no special intervention). Having such information for nonintervention populations (i.e., persons analogous to “controls” in prevention trials) is a critical first step in determining the sample sizes that might be required to detect intervention effects of various magnitudes.
Through medical and administrative records of KPMCP, we retrospectively identified all women who had an STD diagnosed within a 6-month period ending 18 months before the date at which the records were accessed (and before commencement of recruitment for the Psst! trial). We were thus able to follow the medical visits of each member of this cohort over a period of 18 months subsequent to the date of their individual STD diagnoses. The recruitment and follow-up periods for the present cohort preceded and did not overlap those of the Psst! trial.
Our purpose in studying this earlier STD cohort was not to demonstrate already well-established etiologic linkages between STDs and their sequelae but to quantify the rates of utilization for various conditions within a time period commensurate with the follow-up in typical STD intervention trials. We also identified a medical center– and age-matched sample of women seen for some other (non-STD) diagnosis in the same 6-month period as the STD cohort and observed their subsequent 18-month utilization. This comparison group provides one population with which the resource utilization of the STD cohort can be usefully compared. It is not necessarily composed entirely of women who have never had an STD diagnosed.
This study was carried out with the approval of the Institutional Review Boards (IRBs) of the Kaiser Foundation Research Institute and the Palo Alto Medical Foundation Research Institute.
Identification of the STD Cohort
To conduct this retrospective cohort study with a sample selected on the basis of the risk factor, an STD, we identified all KPMCP women members, 18 to 45 years of age, who had an STD diagnosed at the Kaiser Permanente Medical Centers (KPMC) in Oakland, Richmond, or Santa Clara/Milpitas, California, between 1/1/1998 and 6/30/1998. We refer to this as their index visit. A newly diagnosed STD was evidenced by electronic medical record of any one or more of the following: (1) an outpatient medical visit record containing an ICD-9 code for trichomoniasis; (2) a positive laboratory test for chlamydia, gonorrhea (GC), or syphilis; and (3) a positive herpes simplex culture, direct assay, or serology test or a visit record containing an ICD-9 code for genital herpes, unless the patient had any outpatient visit with a herpes code before 1/1/98. A total of 1205 women met one of these three STD criteria.
Relying solely on laboratory evidence of herpes would have eliminated many new diagnoses made by clinicians without laboratory confirmation. Positive laboratory results (direct assay, serology, or culture) identified 195 women; 280 had a herpes diagnosis code associated with their visit; and 395 women met one or both criteria. Of these, 87% had no prior recorded herpes visit, including 91% of those with a positive laboratory test only (n = 115), 91% of those with both a visit code and laboratory confirmation (n = 200), and 82% of those with a visit code but no laboratory confirmation (n = 80). Laboratory confirmation would be very unlikely to be requested if the clinician knew of a prior diagnosis. Including women with no laboratory test result and no prior KPMC herpes diagnostic code might have included some women for whom the present (index) visit was for an exacerbation of herpes diagnosed outside the KPMC system or undiagnosed. At most, this could have affected 164 women (82% of 200); the actual number is almost certainly very much smaller.
Definition of the Comparison Sample
To construct a comparison sample (n = 4820) that was equivalent to the STD cohort in its distribution by medical center and age within medical center, we randomly sampled four times as many women of each age, within each medical center, as were observed in the STD cohort. We drew this sample from among all remaining KPMCP women members, 18 to 45 years of age, who were seen on an outpatient basis between 1/1/98 and 5/31/98 at the three medical centers. Again, their first visit during this period is termed their “index” visit. Temporary unavailability of the June 1998 visit database file prevented sampling of women seen in June 1998 for a first 1998 visit. From the month-by-month pattern of repeated visits of women first seen from January to April 1998, only an estimated 8.1% of all women seen from January to June 1998 would have had their first 1998 visit in June. Group comparisons with women seen for an STD from January to May 1998 did not reveal any substantial differences from the results reported below.
The marginal distributions of the STD and comparison groups by age and medical center are identical by design: 18 to 25 years, 44.1%; 26 to 35 years, 33.9%; 36 to 45 years, 22.0%; mean age = 28.4 years (SD = 7.9); Oakland, 47.1%; Santa Clara/Milpitas, 32.9%; Richmond, 19.9%.
Extraction and Categorization of Subsequent Healthcare Utilization
Types of healthcare utilization.
We were primarily interested in healthcare utilization for STD infections, conditions that may be sequelae of an STD infection, other reproductive health problems, and behavioral/mental health conditions that may be consequences of an STD or of associated relationship issues or that may reflect problems that put women at risk for STDs. We defined 14 diagnostic clusters: (1) STD ruled out; (2) STD; (3) pelvic inflammatory disease (PID) or endometritis; (4) candidiasis; (5) vaginitis; (6) cervical dysplasia; (7) dyspareunia or pelvic pain;(8) disorders of menstruation or abnormal bleeding; (9) contraception; (10) pregnancy: high-risk, complication, or ectopic; (11) infertility; (12) other reproductive health problems; (13) behavioral/mental health conditions; and (14) pregnancy: normal (pregnancy visits by women who had no pregnancy visit coded as high-risk, complicated, or ectopic). The residual (much larger) cluster consisted of all other diagnoses.
For each woman, we extracted from the Outpatient Service Clinical Record database the date of visit, facility, provider, and diagnoses for their index visit and all outpatient clinic visits to any KPMC in Northern California during the 18 months immediately following their index visit. For each office visit, the provider completes a single form on which is recorded all applicable diagnoses. Multiple diagnoses are typical. The index visits were not included in the follow-up observation period.
For the same 18-month time period, we extracted from the Admission-Discharge-Transfer database the date of visit/admission, length of stay, facility, and discharge diagnoses for all inpatient stays (emergency, urgent, elective, delivery) and all outpatient services provided at any KPMCP or KPMCP-contracted hospital or nonhospital facility (outpatient surgery center, hospital ambulatory surgery, and hospital outpatient services). Such utilization does not overlap the office visits described above.
For each diagnostic cluster or combination of clusters, we determined (1) the number and proportion of women who had any visit involving a diagnosis in that cluster and (2) the number of visits at which a diagnosis in a given cluster was recorded (diagnoses recorded on a given visit might be in the same or different clusters). Differences between the STD and comparison cohorts in the probability of having a visit in a given cluster were tested with logistic regression, before and after controlling (where appropriate) for age, medical center of the index visit, pregnancy status, and chronic disease status. Having a chronic disease meant that any of the following were recorded on the patient's “Significant Problems List:” neurotic disorders, COPD or asthma, psychoses, neoplasms, endocrine disease (including diabetes), obesity, hypertension, connective tissue disease, miscellaneous other chronic conditions (e.g., liver disease, cystic fibrosis, and renal failure), thyroid disease, epilepsy, osteodegenerative arthritis, hyperlipidemia, multiple sclerosis, rheumatoid arthritis, ulcerative colitis, CHF/HD, Crohn's disease, and the hemolytic anemias. Because the distributions of the numbers of visits in the various clusters were not normal, STD-comparison group mean differences were evaluated with the Wilcoxon signed-rank test.
Visit rates also were compared, pairwise, between the comparison sample and each of the five individual STD subgroups defined by index STD visit (trichomoniasis, chlamydia, GC, syphilis, and genital herpes), with use of logistic regression. These pairwise comparisons were performed only for diagnostic clusters in which the overall chi-square test value for visit rates among the six groups was statistically significant at the P ≤ 0.05 level. For these analyses, for women who had more than one STD diagnosed on the day of the index visit (n = 33), the index visit was assigned the diagnosis with the lowest overall prevalence in the sample as a whole (see Table 1 and the table footnote). All analyses were carried out with SAS software version 6.12. 14
Trichomoniasis was the most prevalent index diagnosis in the STD cohort, followed by herpes and chlamydia (Table 1). Because 33 women had more than one infection diagnosed at their index visit, the proportions in Table 1 sum to more than 100%. Women with diagnosed GC or chlamydia were significantly younger than women with diagnosed trichomoniasis, herpes, or syphilis.
Outpatient Office Visits
Rates of resource utilization.
Table 2 shows the numbers and proportions of women with one or more visits in a given diagnostic cluster, separately for the STD cohort and comparison sample. Given the large number of diagnostic comparisons between the STD and comparison groups, we adopted a conservative (P ≤ 0.005) alpha level as the criterion for statistical significance. We caution against overinterpreting the comparisons when the p values are greater than 0.005. The definitions of conditions included in each of the clusters are provided in the footnotes to Tables 2–4.
Among women with a diagnosed STD, the likelihood of subsequent visits for reproductive health problems in the follow-up period was substantial and significantly higher than in the comparison group. Relative risks by diagnostic cluster ranged from 3.8 to 2.0 for STDs, pelvic inflammatory disease (PID), visits to rule out an STD, vaginitis, and candidiasis. The rate of candidiasis in the comparison (no index STD) population, 8%, was essentially doubled in the STD cohort, presumably as a consequence of STD treatment.
Other diagnoses for which relative risk was significantly elevated and in the range from 1.7 to 1.3 included cervical dysplasia, disorders of menstruation and abnormal bleeding, ectopic or high-risk pregnancy, pregnancy complications, and behavioral/mental health problems. As footnoted in Table 2, a significantly higher proportion of women in the STD cohort had any pregnancy visit: 0.229 for the STD cohort versus 0.176 for the comparison group (RR = 1.3;P = 0.0001). Among pregnant women (i.e., those who had any pregnancy visit during the follow-up), the relative risk of an ectopic, high-risk, or complicated pregnancy was not significantly higher in the STD cohort than in the comparison cohort (RR = 1.1;P = 0.243). The relative risk of being seen for an apparently normal pregnancy in the STD cohort (i.e., having no visit coded as being for a pregnancy problem [ectopic, high-risk, or complication]) barely missed being significantly elevated by our criterion. The relative risks of women having visits for dyspareunia or pelvic pain, contraception, infertility, and other reproductive health problems were not statistically significant by our criterion. It is important to note that there was no elevated relative risk of visits for all other diagnoses (i.e., nonreproductive and nonbehavioral/mental health) in the STD cohort relative to the comparison cohort.
We examined the degree of overlap among women with various conditions diagnosed during the follow-up period and found that a subset of women (64.5%) accounted for a substantial proportion of the resource utilization (Table 3). Rather than consider all possible combinations of visit diagnoses, for this analysis we combined clusters into five larger groups: (A) STD R/O (ruled out), STDs, PID, and endometritis; (B) candidiasis and vaginitis; (C) cervical dysplasia, dyspareunia, pelvic pain, and abnormal bleeding; (D) any pregnancy problem, infertility, and other reproductive health problems; and (E) behavioral or mental health problems. In the STD cohort, 38.5% of women were seen for a diagnosis in group A and 17.3% for a diagnosis in group B, but only 11.5% were seen for a diagnosis in B who had not been seen for a diagnosis in A. That is, more than three fourths of those seen for a group B diagnosis were also seen for a diagnosis in group A. Similarly, relatively small numbers of women were seen for conditions in groups C, D, and E who had not been seen for a diagnosis in one of the previously listed groups.
The relative risks (STD versus comparison) were highly statistically significant (P = 0.0001), ranging from 2.4 down to 1.4. In the comparison group, the tendency for a core group of women to have multiple reproductive and mental health problems was much lower than in the STD cohort. Only 64% of comparison group women seen for candidiasis or vaginitis (group B) also were seen for a condition in group A, compared with 77% in the STD cohort. Only 76% of comparison group women seen for a group C condition were also seen for a group A or B condition, compared with 92% in the STD cohort. And of those seen for a group E diagnosis, only 75% were seen for any condition in groups A-D, compared with 88% in the STD cohort.
The STD and comparison groups were equivalent, by design, in terms of their distributions by medical center and age within medical center. Chronic disease status was found to be closely associated with both age and medical center, and after matching of these characteristics, the STD and comparison groups were found to be equivalent in the proportion of women whose electronic medical record indicated a significant chronic health problem: 15.9% of the STD cohort and 16.0% of the comparison group. Additional statistical controls for chronic disease status, as well as for age and medical center, tended to slightly increase the relative risk values shown in Tables 2 and 3, with essentially no effect on statistical significance. The STD and comparison groups did not differ in membership turnover patterns. Nearly two thirds of the women in the STD and comparison groups (64% and 66%, respectively) were enrolled in KPMCP for the entire follow-up period, and 80% in each group were enrolled for 12 months or longer.
Proportions of women seen for various diagnoses, as a function of index STD diagnosis.
The STD cohort can be subdivided into five subgroups according to the particular STD diagnosed at the index visit. To determine whether the subsequent pattern of resource utilization differed as a function of the specific index STD diagnosis, we first performed a chi-square or Fisher exact test on the overall differences among six groups (the five STD index diagnosis subgroups and the non-STD comparison group) in the proportions of women with subsequent visits in a given cluster. This test was performed separately for each of the follow-up reproductive behavioral/mental health diagnostic clusters considered previously and also separately for each of the individual infections within the follow-up STD cluster. The overall differences among the six groups proved to be significant at well below the P ≤ 0.05 level for all of the follow-up clusters and STD infections, with the exception of infertility (P = 0.365).
For clusters in which the overall test of group differences was significant, a pairwise test (logistic regression) was performed between the comparison group and each of the five index STD infection subgroups. Requiring that the overall group difference first be significant at the P ≤ 0.05 level gave adequate protection against type I errors in the subsequent pairwise comparisons, which were carried out with and without controlling (statistically) for chronic disease status, age, and medical center. By design, the comparison group was equivalent to the STD cohort as a whole in age and medical center distribution and proved to be equivalent in chronic disease status. However, statistical controls were required because the comparison group was not necessarily comparable to the individual index STD subgroups within the STD cohort. There proved to be very little difference between the controlled and uncontrolled relative risks and P values.
When women were seen during the follow-up for an STD, there was a marked tendency for their STD visits to be for the same STD diagnosis as their index visit (data not shown). For example, 16% of women who had trichomoniasis diagnosed at the index visit were seen again for trichomoniasis during the subsequent 18 months, compared with 1% in the comparison group and 0 to 9% of women with an STD other than trichomoniasis diagnosed at the index visit. Similarly, 5% of those seen for chlamydia at the index visit were subsequently seen for chlamydia, compared with 0.02% of the comparison group women and 0 to 1.3% of those with an STD other than chlamydia diagnosed at the index visit. This tendency to be seen again for the same condition also was marked for those who were seen initially for a new diagnosis of herpes, of whom 13.7% were seen again for herpes in the subsequent 18 months, compared with 0.5% of the comparison sample and 0 to 1.4% of those seen for a diagnosis other than herpes at the index visit. Note that these numbers do not reflect test-of-cure visits, which are not done in this healthcare system except for pregnant women. Single-dose treatment is provided for bacterial infections and is made available for both partners, regardless of the male partner's membership status. However, to the extent they occurred, test-of-cure visits are part of the resource utilization we wished to quantify.
Figure 1 shows the proportions of women in each of the five index STD diagnosis subgroups and in the comparison group who were subsequently seen for PID or endometritis, candidiasis, vaginitis, cervical dysplasia, menstrual disorders or abnormal bleeding, contraception, an abnormal pregnancy, another reproductive health problem, a behavioral/mental health problem, or a normal pregnancy. Relative to women in the comparison group (i.e., those with a non-STD diagnosis), an index diagnosis of GC, chlamydia, or trichomoniasis was associated with a significantly greater likelihood of a diagnosis of PID or endometritis during the follow-up. Index diagnoses of trichomoniasis, herpes, and chlamydia were associated with a significantly greater likelihood of candidiasis and of vaginitis, and a diagnosis of GC was associated with the greatest likelihood of vaginitis. Relative to women with a non-STD diagnosis, cervical dysplasia was significantly more likely in women who had herpes or GC diagnosed at the index visit; menstrual problems and abnormal bleeding were more likely in women with diagnosed trichomoniasis; an abnormal pregnancy was more likely with chlamydia or trichomoniasis; other reproductive health problems were more likely with a diagnosis of trichomoniasis; and behavioral/mental health problems were more likely with diagnoses of trichomoniasis and herpes.
As noted above, a problematic pregnancy was significantly more likely in women with diagnosed chlamydia. When the analysis was restricted to pregnant women, both chlamydia and syphilis were associated with some elevation in the rate of problematic pregnancies, but neither attained the 0.05 level (Pchl = 0.08;Psyph = 0.08). Normal pregnancies were significantly more likely among women with syphilis or chlamydia diagnosed at the index visit than among comparison group women. The reason for this is somewhat unclear but may reflect a combination of the smaller sample size, the low prevalence of syphilis, and the low probability of detecting either condition in the absence of screening tests, which are much more likely to be performed for pregnant women. None of the index STDs was associated with a greater likelihood of being seen for contraception than in the comparison group.
Numbers of outpatient visits.
A further and somewhat independent indicator of resource utilization is the mean number of visits in each diagnostic cluster for women who had any visit in that cluster. Given that at least one visit occurred in a given cluster, the mean number of visits in that cluster typically did not differ significantly between the STD and comparison groups (Table 4). In other words, women in the STD group who developed a given problem typically did not require more visits to address that problem than did women in the comparison group. However, there were exceptions. The mean number of visits to rule out an STD was significantly greater among those seen for that purpose in the STD cohort than among women seen for that purpose in the comparison group. Presumably, this reflects multiple single-visit episodes in the STD cohort. This also was the case for vaginitis. However, STD cohort women who were seen for an STD during the follow-up had approximately 22%fewer follow-up STD visits, per person, than did comparison cohort women seen for an STD during the follow-up. Among women seen for pregnancy problems, infertility, and behavioral/mental health problems, there was a tendency toward higher numbers of such visits in the STD cohort than in the comparison group, but the relative risks did not reach our criterion for statistical significance. Given the greater prevalence of pregnancy and behavioral/mental health problems in the STD cohort, net resource utilization for these problems was still somewhat greater in the STD cohort.
The STD cohort as a whole (i.e., inclusive of women with no follow-up visits at all) averaged a significantly greater number of visits of all types per person in the 18-month follow-up than did the comparison group. The STD cohort had a total of 2533 visits for the various problems listed (exclusive of visits for contraception, normal pregnancy, and behavioral/mental health diagnoses), in contrast with a total of 5518 such visits in the larger comparison sample—an excess of approximately 1 visit per person in the STD cohort (960 excess visits per 1000 women).
The STD and comparison groups did not differ significantly in overall likelihood of inpatient or outpatient hospital care subsequent to their index diagnosis, and they also did not differ in hospital-based care for any of the specific diagnoses investigated above. Approximately 12% of each group had an episode of hospital-based care, 8% had at least one inpatient admission, and 5% were seen at least once by a hospital-based outpatient service. The STD cohort averaged 3.28 (SD = 1.86) days of inpatient care for Ob-Gyn or behavioral/mental health problems and the comparison sample averaged 3.27 (SD = 1.86).
Following diagnosis with an STD, 16.4% of women were seen for another STD, and a total of 38.5% were seen for an STD, for PID, for endometritis, and/or to rule out an STD. That this occurs is hardly surprising. However, the full extent of near-term resource utilization for known sequelae of STDs may not be fully appreciated. Over half (57.6%) of these women were seen for at least one (and typically several) reproductive health problems. These results, derived from a closed panel healthcare system, provide better quantitative estimates of resource utilization subsequent to an STD than do data from community samples receiving care from multiple sources.
The pattern of resource utilization among women with a diagnosed STD differs substantially from that of women seen for other problems. Women with a diagnosed STD are, variously, from two to four times more likely than the comparison women to be seen for evaluation of a possible STD or to have an STD diagnosed; two to four times more likely to be seen for candidiasis, vaginitis, PID, and/or endometritis; and also significantly more likely to be seen for cervical dysplasia, disorders of menstruation and/or abnormal bleeding, a high-risk, ectopic, and/or complicated pregnancy, and a variety of behavioral and mental health problems (notably, counseling and education about HIV/sexual issues).
Limitations and Potential Confounders
The differences we observed between the STD and comparison cohorts are most likely underestimates of the difference in utilization of reproductive health services one might expect between women who ever and those who never experience an STD infection. The fact that 35.4% of women in the comparison group experienced one or more of the Ob-Gyn problems examined here suggests that some of these women had a history of STD infections, whether ever diagnosed or not. However, we did not actually characterize prior healthcare utilization for STDs and associated conditions in our population. Furthermore, differential utilization in 18 months may not be reflective of differential utilization over a longer time span.
Ethnicity and marital status.
We interpret the observed significant differences between the office-based healthcare utilization of the STD and comparison groups in the follow-up period as being the result of the index and/or previous STD infections. This interpretation clearly is biologically plausible. However, unmeasured differences between the groups could conceivably account for a portion of the observed differences. KPMCP records do not contain information on patient ethnicity or marital status, which are known to be associated with STD incidence 14; hence, it was not possible to control for these individual characteristics statistically or by design.
The triennial KPHP member survey conducted in spring 1999 documents that the Oakland and Richmond medical centers have higher proportions of black and lower proportions of white non-Latino/Hispanic patients than does the Santa Clara/Milpitas center, which has higher proportions of Latino/Hispanic and Asian patients. STD rates are higher at the Oakland and Richmond sites. Despite matching the STD and comparison groups on medical center, it is likely that the STD cohort still includes a higher proportion of black, lower-income/education, and single women than does the comparison sample. If so, the observed differences in resource utilization conceivably might result from other factors that are associated with ethnicity, income, education, or marital status, beyond their association with having had an STD. This would be the case if, for example, an overrepresented subgroup in the STD cohort also had a generally greater (or lesser) tendency to seek medical care for any given set of symptoms.
If such differences in care-seeking behavior did exist, a more accurate estimate of the impact of STDs would be obtained by comparing STD and non-STD samples within a given ethnic, income/education, or marital status subgroup, but this is not possible in the current study, and we did not seek to establish ethnicity- or marital status–specific utilization rates. That a substantial portion of the observed difference is a biologic consequence of STD infections and their treatment and not an artifact of unmeasured cohort differences is suggested by the fact that the relative risk was highest and statistically significant in those diagnostic clusters most clearly STD-related, while not elevated (RR = 1.0) for diagnoses having no relationship to reproductive or behavioral/mental health (i.e., the cluster “All other diagnoses”).
Infertility and pregnancy problems.
Despite the established etiologic relationship between N gonorrhoeae and C trachomatis infections and PID, and between PID and infertility, as well as associations between STD infections and other pregnancy problems, we observed no statistically significant difference in the rates of infertility or of problematic pregnancies between the STD and comparison cohorts when the comparison was restricted to pregnant women. A significant difference in overall problematic pregnancy rates was observed, however, when all the women were considered, without regard to pregnancy status. When women with different types of index STD infections were considered, it was those with diagnosed chlamydia or trichomoniasis who differed most from the comparison group in their rate of problematic pregnancies, although these differences were not statistically significant. This suggests that one factor contributing to the significant difference when all women are considered may be the higher overall pregnancy rate in the STD cohort. In addition, the comparison restricted to pregnant women involves smaller samples. Finally, the subgroups of pregnant women are not matched by design, unlike the STD and comparison groups as a whole.
Implications for STD/HIV Prevention
The association documented here between an STD diagnosis and relatively high rates of resource utilization provides additional justification, should any be needed, for aggressive efforts to prevent sexually transmitted infections and for additional efforts to design effective interventions, including interventions targeting managed care populations. During an 18-month period, approximately 30% of the women with a diagnosed STD were seen again to rule out an STD, for at least one subsequent STD, and/or for PID or endometritis. Twenty-six percent were seen for behavioral or mental health problems. Twenty-eight percent of those who became pregnant experienced some problem with that pregnancy. Furthermore, depending on which problems are included, a subset consisting of approximately two thirds of the women seen for an STD were responsible for the majority of the observed reproductive and mental health morbidity and consequently for the very substantial resource utilization and cost of care.
PID, ectopic and other problematic pregnancies, and infertility are especially costly conditions to treat. A single index STD identified approximately 20% to 25% more women who would experience such problems than among women who were not seen for an STD in the same 6-month period (an undetermined proportion of whom may have had a prior STD), and women seen for an STD proved to be an average of 1.5 to 2 times more likely to be seen subsequently for these various problems than women not seen for an STD.
Targeting STD prevention interventions early enough to avert even a first infection and certainly a second should have the greatest potential benefit, especially for girls who may be at risk for engaging in unsafe sexual practices. However, intervention arguably should be undertaken whenever an STD is diagnosed, including at the first positive diagnosis from a screening test, such as for chlamydia. It appears especially important to better characterize the subset of women most likely to experience repeated STDs and their sequelae, in order to facilitate intervention before the establishment of this pattern.
N gonorrhoeae and C trachomatis infection play an etiologic role in PID in the US. 3 Intervention to screen for and treat chlamydia has been shown to reduce the incidence of PID and to be cost-effective in appropriate populations. 2,15,16 Chlamydia screening is currently recommended for all women ≤25 years of age and all women with mucopurulent cervicitis and specific sexual behavior risk characteristics. 17 However, at present, intervention to prevent future infection typically is not undertaken in a systematic manner upon diagnosis of an STD in a managed care setting. Such interventions are not cost-free, and their effectiveness in preventing further STDs remains to be adequately demonstrated in such settings. Even if well-planned interventions can reduce the risk of subsequent STD infections, the extent to which they can reduce STD sequelae and complications, and over what time course, is as yet unclear. Such benefits may be limited by the extent to which substantial and irreversible damage has already occurred, again reinforcing the need to intervene as early as possible in women's reproductive careers.
Application in Conducting Health Services Research and STD Prevention Trials
The methods and results of the current study can assist in planning sample size and data analysis strategies, including cost effectiveness/cost benefit analyses, in prevention intervention trials targeting women who have experienced an STD and in other related health services research—research clearly warranted by the present results. Such research requires identification and classification of relevant utilization of services. As part of the methodology of the current study, we conducted a very careful analysis of all ICD-9 codes to identify those that were relevant and then grouped them in such a way as to reduce the detail involved and still preserve distinctions of interest. This detailed categorization of ICD-9 diagnostic codes can be useful in the needed research.
Sample size determination requires an estimate of the rates of relevant outcomes or counts of relevant events under different conditions in order to estimate treatment effect sizes (e.g., differences in outcomes between intervention and nonintervention (control) groups in randomized clinical trials). For example, consider an STD prevention intervention trial targeting women who have previously had an STD diagnosed in which a planned outcome measure is the proportion of women with a visit for any STD-related or other relevant reproductive health problem during the follow-up. Among women with a diagnosed STD for whom no intervention is initiated (the control group), the present results suggest that 57.4% might be expected to have one or more visits for an STD-related condition over the succeeding 18 months. Assuming the intervention were to reduce the proportions experiencing such problems to the level of the present comparison sample (35.1%), the sample size required in order to detect such a difference with a given level of power and type I error rate can be determined. Similar calculations can be carried out with assumptions of any desired intermediate effect size and for other specific outcomes, such as STD visits, with use of the present results.
Women with a diagnosed STD are at substantially increased near-term risk, not only of subsequent STDs but of being seen in a healthcare setting for other serious, costly, and potentially life-threatening reproductive health problems. This study has quantified the elevated resource utilization risk for adult female members in an ethnically diverse managed healthcare system, as manifest within a period of 18 months following an index STD diagnosis. The results underscore the need for more systematic efforts to target such women with effective STD prevention programs, both to improve individual and public health and to reduce healthcare costs. These results also can be useful in planning both sample size and data analysis strategies in STD prevention interventions, especially those targeting women with a diagnosed STD.