Incidence Rates of Pelvic Inflammatory Disease Diagnoses Among Army and Navy Recruits: Potential Impacts of Chlamydia Screening Policies

Introduction 

Pelvic inflammatory disease (PID) is a potentially severe condition of the female genital tract resulting from the ascent of pathogenic micro-organisms from the lower to the upper reproductive organs.1 Case presentations vary, from symptomless to symptoms of lower abdominal pain, fever, abnormal vaginal discharge or bleeding, dyspareunia, and dysuria.2 Diagnosis is frequently based on clinical findings, including lower abdominal or pelvic pain with bilateral adnexal, uterine, or cervical motion tenderness without competing diagnoses.3 Acute and chronic sequelae may result from untreated, unsuccessfully treated, or recurrent bouts of PID, and may include chronic pelvic pain, ectopic pregnancy, and tubal infertility.4, 5

Chlamydia trachomatis is recognized widely as the most common sexually transmitted bacterial infection in the U.S.6 as well as a common cause of PID.1 Risk factors for genital chlamydia infection and PID are similar including age <257; African-American race/ethnicity8; recent new sexual partner(s), multiple sexual partners, or both9; lower education10; choice of contraception11; and southeastern U.S. provenance.8 Frequent asymptomatic clinical expression (70%–90%)12 and the progression of untreated infection to PID (40%)13, 14 implicate chlamydia as a substantial threat to women's health.

United States military women constitute a high-risk population for chlamydia infection. In one study, age- and race-adjusted infection rates among symptomatic women in the U.S. Army were six times higher than among their counterparts in the U.S. population.15 A recent screening study reported asymptomatic chlamydia infection in 9.5% of female Army recruits,16 consistent with earlier reports.17, 18 In another recent report, the prevalence of asymptomatic infections among a representative sample of U.S. women aged 18–26 years was 4.7%.8 Female chlamydia infection rates appear to be on the rise among both military recruits16 and the general U.S. population.3

In response to the threat posed by chlamydia to young women in military service, in May 1999 the U.S. Armed Forces Epidemiological Board (AFEB) recommended the screening of all female recruits soon after accession, followed by annual screening.19 Screening within 1 year of accession during the routine annual gynecologic exam, the policy adopted by the U.S. Army,20 was considered acceptable and in compliance with recommendations made by the U.S. Preventive Services Task Force.21 In contrast, the U.S. Navy has required chlamydia and gonorrhea screening of all female recruits within a few days of accession,22 followed by annual exams until age 25. The Navy's screening of new recruits is combined with an information program on sexually transmitted infections.

It was hypothesized that the differences in service-specific chlamydia screening policies, which may result in delays of up to 12 months for the diagnosis and treatment of infections following accession into the Army, would result in a greater rate of PID among Army compared to Navy accessions. For this report, the records of medical encounters routinely collected for public health surveillance purposes were used to assess the differences in incidence rates of outpatient diagnoses of PID. A retrospective cohort of female accessions to the Army and Navy during a 5-year surveillance period was considered.

Results 

A total of 60,268 Army (63.8%) and 34,173 Navy (36.2%) accessions fulfilling the inclusion criteria were initially identified in the DMSS. Following the exclusion of service members with missing covariate data, 58,088 Army (63.7%) and 33,046 Navy (36.3%) accessions—contributing 93,918.14 and 65,863.39 person-years at risk, respectively—constituted the study sample (Table 1, Table 2). During a median (range) follow-up time of 18.04 (0.03–59.99) months, 1276 Army and 546 Navy outpatient PID events were identified. The crude incidence rates of 13.59 cases/1000 person-years and 8.29 cases/1000 person-years were observed for the Army and Navy, respectively. The rate ratio (95% CI) for Army versus Navy was 1.64 (1.48–1.81; p<0.0001).

Table 1. Demographic characteristics of women accessioned to active military service between January 1, 2001, and December 31, 2005 (N=91,134)

	Factor	Army, n (%)	Navy, n (%)
	Service	58,088(63.7)	33,046(36.3)
	Mean (SD) birth year⁎	1982.7(2.4years)	1983.1(2.2years)
	Mean (SD) accession year⁎	2002.9(1.4years)	2002.8(1.4years)
	Race/ethnicity⁎
	White	30,760(53.0)	17,947(54.3)
	African American	16,058(27.6)	7,720(23.4)
	Hispanic	7,737(13.3)	3,107(9.4)
	Asian	2,426(4.2)	1,734(5.2)
	Native American	986(1.7)	2,038(6.2)
	Other	121(0.2)	500(1.5)
	Home-of-record statea,⁎
	Other than Southeast	34,120(58.7)	20,560(62.2)
	Southeast	23,968(41.3)	12,486(37.8)
	Education levelb,⁎
	<High school	120(0.2)	312(0.9)
	High school	50,912(87.6)	30,298(91.7)
	Some college	2,132(3.7)	953(2.9)
	Bachelor's	4,852(8.4)	1,474(4.5)
	Master's	39(0.1)	6(0.0)
	Doctorate	33(0.1)	3(0.0)
	Marital statusb,⁎
	Unmarriedc	50,458(86.9)	31,990(96.8)
	Married	7,630(13.1)	1,056(3.2)
	Rankb,⁎
	E1–E4	54,630(94.0)	31,912(96.6)
	E5–E6	32(0.1)	11(0.0)
	O1–O3d	3,426(5.9)	1,123(3.4)

a Southeast indicates state reported as AL, DC, DE, FL, GA, KY, MD, MS, NC, OK, SC, TN, TX, VA, or WV b Value at time of accession c Includes divorced and widowed d Army includes W1 and W2 ⁎ p<0.0001 for difference in distributions between Army and Navy

Table 2. Incidence of outpatient pelvic inflammatory disease (PID) diagnosis among women accessioned to active military service between January 1, 2001, and December 31, 2005 (N=91,134)

	Factor	Army: cases	Person-yearsa	Navy: cases	Person-yearsa	Incidence rateb	Crude rate ratio (95% CI)	Adjusted rate ratioc (95% CI)
	Service	1276	93,918.1	546	65,863.4	11.40	1.64(1.48–1.81)	1.62(1.46–1.79)
	Birth yeard	281	20,131.3	104	10,787.6	12.45	1.01(0.99–1.03)	n/r
	Accession yeare	218	18,933.5	78	11,630.7	9.68	1.00(0.96–1.04)	n/r
	Race/ethnicity
	White	537	46,122.3	262	34,342.1	9.93	ref	ref
	African American	522	28,167.0	174	15,505.3	15.94	1.60(1.45–1.78)	1.50(1.36–1.66)
	Hispanic	171	13,704.8	59	8,624.8	10.30	1.04(0.90–1.20)	0.97(0.84–1.12)
	Asian	31	4,258.0	19	3,425.3	6.51	0.66(0.49–0.87)	0.67(0.50–0.89)
	Native American	11	1,568.5	27	3,439.6	7.59	0.76(0.55–1.06)	0.82(0.59–1.14)
	Other	4	97.6	5	526.2	14.43	1.45(0.75–2.80)	1.72(0.89–3.32)
	Home-of-record statef
	Other than Southeast	708	55,267.9	306	41,531.7	10.48	ref	n/r
	Southeast	568	38,650.3	240	24,331.7	12.83	1.22(1.12–1.34)
	Education levelg
	<High school	9	582.4	7	648.4	13.00	0.67(0.60–0.75)h	n/r
	High school	1177	82,239.0	517	60,615.6	11.86
	Some college	59	2,805.3	15	1,656.2	16.59
	Bachelor's	31	8,208.6	7	2,915.9	3.42
	Master's	0	55.4	0	14.6	0.00
	Doctorate	0	27.3	0	12.7	0.00
	Marital statusg
	Unmarried	942	72,922.6	406	52,468.4	10.75	ref	ref
	Married	334	20,995.5	140	13,395.0	13.78	1.27(1.14–1.40)	1.31(1.18–1.46)
	Rankg
	E1–E4	1235	84,753.9	531	61,520.7	12.07	ref	ref
	E5–E6	18	2,705.9	11	1,884.3	6.32	0.52(0.36–0.76)	0.50(0.35–0.72)
	O1–O3i	23	6,458.4	4	2,458.4	3.03	0.25(0.17–0.37)	0.25(0.17–0.37)

n/r, not retained

a Person-years may not sum to total due to rounding error. b Crude incidence rate per 1,000 person-years at risk at listed covariate value c Adjusted for all other variables retained in the adjusted multiple Poisson regression model d Case count, person-years at risk, and crude incidence rate for median value equal to 1983 for Army and Navy e Case count, person-years at risk, and crude incidence rate for median value equal to 2003 for Army and Navy f Southeast indicates state reported as AL, DC, DE, FL, GA, KY, MD, MS, NC, OK, SC, TN, TX, VA, or WV g Case count, person-years at risk, and incidence rate presented for covariate values at the time of accession h Education level treated as an ordinal variable i Army includes W1 and W2

Accessions to the Army were more likely to be African American or Hispanic, less likely to be Asian or Native American, more likely to be from the southeastern U.S., and more likely to have a post-secondary education, be married, and have higher rank compared to the Navy (Table 1). Service was significantly associated with each covariate (p<0.0001). In crude Poisson regression models (Table 2), Army service (rate ratio[RR]=1.64 versus Navy); African-American race/ethnicity (RR=1.60 versus white); home of record from the southeast U.S. (RR=1.22 versus non-southeast U.S.); education (RR=0.67 per education level); married (RR=1.27 versus unmarried); and rank (RRE5–E6=0.52; RR01–03=0.25 versus E1–E4) were significant predictors for PID.

The adjusted Poisson regression model comprised service (RRArmy=1.62; 95% CI=1.46–1.79); race/ethnicity (RRAfrican-American=1.50; 95% CI=1.36–1.66; RRAsian=0.67; 95% CI=0.50–0.89); marital status (RRMarried=1.31; 95% CI=1.18–1.46); and rank (RRE5–E6=0.50; 95% CI=0.35–0.72; RR01–03=0.25; 95% CI=0.17–0.37) as significant predictors for PID (Table 2). Neither home of record nor education level was significant in the multivariate model. No covariates substantially changed the estimated RR for Army versus Navy, nor were significant interactions observed.

Time-to-event functions estimated by service are presented in Figure 1. Differences in time to event, by service, began during the 4–8-month follow-up interval, in which Army women experienced a greater risk for PID diagnosis than Navy women. During the 12–16-month interval, a substantial increase in risk was observed among Army accessions, and this difference continued to accumulate through the remainder of follow-up. Time to event was significantly different by service (p<0.0001).

View Large Image Download to PowerPoint

Figure 1. Cumulative risk (%) for an incident outpatient pelvic inflammatory disease (PID) diagnosis among women accessioned to active military service between January 1, 2001, and December 31, 2005 (N=91,134).

Instantaneous hazard functions are presented in Figure 2. An initial increase in risk for Army subjects during the 8–12-month follow-up interval preceded a long and irregular decline, becoming statistically indistinguishable from Navy subjects during the 24–28-month interval and thereafter. In contrast, the hazard function for Navy women remained fairly uniform over much of the duration of follow-up. No overlap of service-specific 95% CIs for the hazard functions was observed, beginning around 8–12 months of follow-up and continuing through 20–24 months, indicating that differences in risk were significant. A substantial overlap of the 95% CIs for service-specific hazard rates was observed after the 24th month of follow-up.

View Large Image Download to PowerPoint

Figure 2. Instantaneous risk (95% CI) for an incident outpatient pelvic inflammatory disease (PID) diagnosis among women accessioned to active military service between January 1, 2001, and December 31, 2005 (N=91,134).

Discussion 

Crude incident PID rates of 13.6 cases per 1,000 person-years at risk for the Army and 8.3 cases per 1000 person-years for the Navy were observed among women aged <25 who accessioned into active military service between January 1, 2001, and December 31, 2005. These observations are consistent with those reported from a clinical trial of high-risk members of a Seattle, Washington–based HMO in which women were randomized to groups for usual care (n=1598) or chlamydia screening (n=1009), and followed for PID for 12 months.28 The study reported 21.7 cases per 1000 person-years at risk among women receiving usual care, and 9.3 cases per 1000 person-years among the screened women. A crude PID RR (95% CI) equal to 2.27 (1.11–5.00) was reported for the usual-care group versus the screening group, and is consistent with that observed for the Army versus the Navy in the current study, which was 1.64 (1.48–1.81). This difference was not explained by age, marital status, race/ethnicity, or education in either study.

Associations observed for the covariates considered in the current study—including age, time in service, race/ethnicity, home of record, education, and rank—were generally consistent with those previously reported for chlamydia and/or PID.8, 29 The observed 31%-increased adjusted risk for PID among married compared to unmarried women (Table 2) was unexpected, given the reported associations among multiple sex partners, new sex partners, or both, and chlamydia infection among female military accessions.17 The authors are unable to explain this observation. Case ascertainment bias may explain, at least in part, the apparent increased risk for PID in married versus unmarried women. For example, while assessing female-specific healthcare utilization practices, the authors documented higher rates (RRMarried=1.20; 95% CI=1.18–1.22) of diagnoses of other disorders of the female genital tract (ICD-9 617–629) among married versus unmarried female service members (data not shown).

Time-to event-analysis suggested that the increased risk for PID observed among Army accessions occurred primarily between 8 and 16 months post-accession, and persisted through the 24th month of follow-up (Figure 2). Variation in the RR for Army versus Navy was attributable primarily to variation in the PID rate among Army women, as the hazard rate for the Navy remained fairly uniform over the duration of follow-up. The authors a priori anticipated that a service-specific difference in PID risk would be evident in approximately the first 12 months following accession. The extended period of increased risk observed for the Army may have been due in part to high recurrence rates for chlamydia among young Army women,30 in conjunction with a positive dose–response relation between chlamydia infection and PID.31 Furthermore, the low rate of reported compliance with current policies regarding annual screening for chlamydia in the U.S. military (∼35–40%) may have contributed.32

Several advantages were offered by the current study. However, a series of limitations were introduced by the complex pathophysiology of PID and the use of routinely collected surveillance, rather than hypothesis-targeted data. Advantages included a large sample of the at-risk population, a large number of captured outpatient PID events, and the relatively complete characterization of selected covariates. Limitations were related to the multi-causal nature of PID, case and exposure ascertainment, and limited covariate availability.

Other organisms, such as Neisseria gonorrhea or those associated with bacterial vaginosis, may have caused the captured PID events, thus biasing study results. Chlamydia and gonorrhea, however, account for a substantial number of PID33 cases, and in nonmilitary populations, approximately 70% of women asymptomatically infected by the latter have been found to be co-infected by the former.8 Furthermore, cross-sectional surveys among female Navy personnel suggest the scarcity of asymptomatic N. gonorrhea infection.34 It is possible that some percentage of captured PID events was associated with gonorrhea, but this was likely small.

Case ascertainment was based on disposition diagnosis only (with no requirement for laparoscopic confirmation), raising the specter of case-misclassification.35 In addition, only incident cases available in the DMSS were captured. Given the high reported recurrence rate of PID (∼14.2%),11 there was a possibility that the PID events captured from the DMSS as incident were not true incident events for a subject. Differential distribution of such misidentified incident cases by service might have increased the baseline PID risk for that population and thus biased the study's results; however, there was no reason for this to have been expected.

Case-reporting bias, the possibility that Army events are reported to the DMSS more frequently than Navy events because of the nonroutine capture of shipboard medical data, was also of concern. An unrelated diagnosis for which service-specific differences in incidence rates were not anticipated—pneumonia and influenza (PNI, ICD-9 480–487)—was used to evaluate this concern. An RR of 0.8 for Army versus Navy, substantially different from that observed for PID, suggested that systematic case-reporting bias did not completely account for the observed service-specific differences in PID rates. Furthermore, the PNI rate ratio varied considerably by year of accession, further undermining the threat of a systematic bias in the reporting of outpatient Army and Navy events (data not shown).

Exposure assessment in the current study was limited. Baseline laboratory data to assess the prevalence of sexually transmitted infections in new female recruits were not available in the DMSS at the time of study. This necessitated the use of a proxy variable, service, as a marker for chlamydia screening. This approach was subject to bias and misclassification. Few data on chlamydia prevalence have been recently published. The most recent data, from a study of female Army recruits during 1996–1999, reported a crude, overall prevalence of chlamydia infections of 9.5% using a nucleic acid amplification test. Substantial variation by race was found.16 A lower crude prevalence, 4.3%, was reported for the female Navy recruit screening program between 1997 and 1999.22 The difference in estimates, at least in part, is probably due to the lower sensitivity of the unamplified test used by the Navy.22, 36 Furthermore, confounding by racial differences in the female recruit populations may also have played a role. It is our opinion that if the effects of these differences were taken into accounted, it is likely that baseline prevalences of chlamydia by service would be similar. In addition, data regarding sexual behavior to include the number and frequency of new partners were unavailable.17 Thus, the possibility of unmeasured confounding by these factors cannot be ruled out.

The current study was not an analysis per se of the effect of chlamydia screening on PID incidence, and the observed difference in PID risk between the Army and Navy cannot be attributed to service-specific implementation of the 1999 AFEB chlamydia screening recommendation for new accessions. The presence or absence of the initial-entry screening of female recruits for chlamydia and gonorrhea, the completeness and effectiveness of annual screening, and the presence or absence of information and education programs are all variables that could have had some impact on the results of this study. Nevertheless, the Navy policy of screening new accessions almost immediately may in part explain the reduced long-term risk for PID observed in that service compared to the Army, which only requires new accessions to be screened within a 12-month time frame.

Despite several methodologic limitations posed by the use of surveillance data, the observed 62% adjusted increase in PID risk for Army compared to Navy accessions is not likely due entirely to confounding or bias, and is consistent with results published from a clinical trial among high-risk civilian women.28 The prevention of chlamydia infections and PID are important to U.S. forces. Military members are often deployed to remote areas where medical services may be limited. Clinical PID may not only compromise the uniformed member's ability to perform the military mission, but also could require the uniformed member's costly evacuation to more definitive care and replacement.37 The results of this study warrant the design of a future confirmatory study to address the limitations of the current framework and to focus more directly on the source of the increased risk for PID observed among Army compared to Navy accessions.