Pelvic inflammatory disease (PID) comprises a range of upper genital tract inflammatory disorders in women that usually result from microorganisms ascending from the cervix to the upper genital tract. PID can result in a wide range of symptoms and no sign or symptom is pathognomonic for the disease.1,2 Therefore, PID is difficult to define as well as to diagnose; and a clear, consistent clinical case definition for surveillance purposes is lacking.3 A standardized methodology for identifying and monitoring PID diagnoses is required.
Understanding the epidemiology of PID is of importance not only in itself, but because the condition is associated with reproductive morbidity, including ectopic pregnancy and tubal factor infertility.4,5 Furthermore, a key aim of chlamydia screening, which is becoming widespread in developing countries, including England, is to prevent the complications of untreated chlamydia, including PID.6–8 Although there are well-documented difficulties in determining the precise proportion of PID that is attributable to genital chlamydia infection, the available evidence suggests that it is substantial.3 Monitoring the frequency of PID is therefore undoubtedly important to inform and evaluate the impact of chlamydia screening.
The United Kingdom General Practice Research Database (GPRD) is the world's largest quality assured database of anonymized longitudinal medical records from general practice. It is held by the GPRD Division of the Medicines and Healthcare Regulatory Agency who regularly update and validate the data. The dataset covers approximately 5.5% of the UK population, and is broadly demographically representative.9 The GPRD has previously been used to examine trends in the incidence of chlamydia diagnoses.10,11 Since the majority of PID cases are likely to present in the general practice setting rather than in hospitals,12 the GPRD provides an ideal dataset for investigating trends in PID diagnoses. We used the GPRD to estimate the frequency of clinical PID diagnosed in general practice in England, and investigated variations by definition of PID, time, age, and geographical area.
MATERIALS AND METHODS
We obtained an extract of the GPRD from the Medicines and Healthcare Regulatory Agency for the years 2000 to 2008. Data for 2008 were only available for the first 6 months of the year but were considered adequately representative of the full year. Diagnoses are recorded in the GPRD data using “Read” or “Oxford Medical Information System (OXMIS)” codes (each has a corresponding GPRD medical code) assigned during clinical consultations. Each consultation is recorded as a single record (row of data) in the GPRD dataset that contains a “Read” or “OXMIS” code, along with other demographic and clinical information. The GPRD extract used in our analyses consisted of all GPRD records for female patients aged 16 to 44 years containing any one of a defined list of “Read” or “OXMIS” codes, denoting PID diagnoses or symptoms indicative of acute or chronic PID. We subclassified this list into codes for “definite,” “probable,” and “possible” PID cases (Supplemental Digital Content, Appendix 1, online only, available at: http://links.lww.com/OLQ/A9). Diagnoses or symptoms of PID related to childbirth or abortive pregnancy were excluded (but not all cases occurring during pregnancy). All codes denoting chronic cases were excluded from the “definite” category and instead classified as either “probable” or “possible,” as appropriate. Only data from practices complying with specific quality measures defined by the GPRD, based on the overall consistency and completeness of the data provided by the practice, were included. Approval for this study was given by the GPRD Independent Scientific Advisory Committee.
A woman may have more than one record (PID event) in the dataset—multiple records for the same woman are identified using a unique GPRD patient number. Events occurring within 42 days of each “index” PID event for a patient were considered as the same, single episode of PID, which was based on genitourinary medicine (GUM) clinic practice for dealing with a sexually transmitted infection episode. We investigated the sensitivity of this 42-day case definition by calculating the rates using 60- and 90-day definitions. As this had very little effect on rates (less than 2% lower rates of definite/probable PID using a 90-day definition), we used the 42-day definition in all analyses. Each episode was then classified as “definite,” “probable,” or “possible” PID in that order of priority, that is, if there were any “definite” codes during an episode, the episode was classified as “definite,” whereas an episode was only classified as “possible” if there were no “definite” or “probable” codes during the episode. For most of the analyses, a combined “definite/probable” category was used.
Aggregated denominator data on person-time among the 16- to 44-year-old female population registered in the GPRD were used to calculate PID rates, both for overall and for each subgroup analyzed. The inclusion of person-time for females who had PID, and were therefore not susceptible to PID, was considered a negligible error in the denominator. Incidence rates are presented per 100,000 person-years (py), with 95% confidence intervals (CI).
Trends were assessed using Poisson regression, and percentage changes in PID incidence over time were measured on the log scale. P values were obtained using the Wald test. Data were analyzed using STATA 10.0.
Absolute rates of PID varied considerably with the definition used; the overall rate of definite PID among 16- to 44-year-old females was 281/100,000 py (95% CI: 277–286), 326/100,000 py (95% CI: 321–331) for definite plus probable cases, and increased to 1117/100,000 py (95% CI: 1107–1126) when possible cases were also included. There were a total of 14,873 episodes of definite/probable PID during the study period, 10.8% were repeat episodes (13,260 women had PID).
There was heterogeneity in the rates of PID by age group and region. The rate of definite/probable PID cases was highest among women aged 20 to 24 years (505/100,000 py; 95% CI: 488–522), followed by those aged 25 to 29 years (405/100,000 py; 95% CI: 391–420). The rate remained highest in these 2 age groups when possible PID cases were also included (Table 1).
During 2000 to 2008, the rate of definite PID decreased by 10.2% per year (95% CI: 9.5–11.0; P < 0.001) and probable PID cases by 11.4% per year (95% CI: 9.6–13.2; P < 0.001). In contrast, the rate of possible PID cases increased by 2.5% per year (95% CI: 2.0–2.9; P < 0.001), largely due to increase in episodes of pelvic pain. When definite, probable, and possible cases were combined, the rate declined by just 1.2% per year (95% CI: 0.9–1.6; P < 0.001) (Fig. 1).
Further analyses were conducted using the definite/probable case definition only. During 2000 to 2008, the rate of definite/probable PID cases decreased by 10.4% per year (95% CI: 9.7–11.1; P < 0.001). A significant (P < 0.001) decline was apparent in all age groups and was greatest among 16 to 19 year olds (12.7% per year; 95% CI: 10.8–14.7; P < 0.001) (Fig. 2).
The rate of PID diagnoses declined in all areas of England (although the decline was nonsignificant in the North East) during 2000 to 2008. Declines were greatest in the West Midlands (12.5% per year; 95% CI: 10.6–14.4; P < 0.001) and the North West (12.1% per year; 95% CI: 10.6–13.7; P < 0.001).
The GPRD dataset offers an excellent opportunity to investigate and monitor trends in the rate of clinical PID. Our study demonstrates that the specificity of PID definition used has a major effect on the absolute rates, and also on trends over time. There was a decline in definite/probable PID diagnosis rates during the study period, for England as a whole, in all areas, and in all age groups. Cases meeting the possible case definition increased over this time.
The major strength of this study is that it is based on a large, national sample of general practice patients that is representative of the general population. The participation of practices in GPRD is fairly stable, and the representativeness of the participants is reviewed regularly. In addition, as PID is largely diagnosed in general practice setting, this dataset should provide a better estimate of the burden of PID in the population than other sources such as hospital data.3,12 There are, however, also some limitations. First, the study includes women with clinical PID only (i.e., those presenting to healthcare services). Because PID can be mild or asymptomatic,2 the true frequency of the disease among the population may have been underestimated. This should not, however, affect the validity of the observed trends over time. Second, the diagnosis of PID is not straightforward—laparoscopy is considered the “gold standard” but is of limited use due to its costly, invasive and, in many settings, impractical nature.3 A clinical PID definition was therefore used in this study, as in others,12,13 although it is recognized that the symptoms of PID can be nonspecific and thus lack sensitivity and specificity as a means of diagnosis.14–16
The reason for the increasing rate of possible PID (comprised largely of patients presenting with pelvic pain) over the study period is not clear. Chronic pelvic pain is a common diagnosis in general practice, often without an identified underlying cause,17 and therefore, probably a poor marker of clinical PID. The definite/probable definition used in this study is the most comparable (though not identical) to the International Classification of Diseases (ICD) coding of PID, that is, excluding vague symptoms such as pelvic pain,18 and likely gives a better measure of the true rate of PID diagnoses and associated trends in general practice. Two earlier studies of PID incidence in UK general practice using ICD9 codes to define PID cases also demonstrated declines in PID incidence over time. However, the absolute rates reported in these studies were far higher than the rate of definite/probable PID in our study (1670 per 100,000 py, among women aged 16 to 46 years during 1991 to 1992,12 and 1787 per 100,000 py among 15 to 44 year olds during 1994 to 2001,13 compared to 326/100,000 py among 16 to 44 year olds in our study). Some of these differences may be explained by the continuation of the decline in the rate of PID reported in the earlier studies, and possible different methods of defining PID episodes (e.g., how multiple episodes were dealt with). In addition, differences in the definitions of PID will have had an impact. The latter study13 included ICD codes for inflammatory disease of the cervix, vagina, and vulva (ICD9 616) that were not included in our code list because these are not inflammatory conditions of the upper genital tract. A reanalysis of the database used for this previous study (the Royal College of General Practitioners Weekly Returns Service) for the year 2001, has shown that over 70% of episodes coded to ICD9 614 to 616 were in fact ICD9 616, suggesting that this may account for most of the difference between the frequency of PID found in that study and ours, for the years in common to both (M. Barley and R. Pnaiser, personal communication, 2010). Comparison with the other previous study,12 which defined cases of PID by ICD9 code 614 only, is complicated by the much earlier period and the different data source—the study was based on a relatively small survey of general practices in England and Wales (60, representing approximately 1% of the population). Despite the differing absolute rates between these studies and ours, all provide consistent evidence that diagnoses of PID are in decline.
The proportion of women with symptoms of PID presenting at general practice rather than other clinical settings may have changed over time. A recent study has shown that an increasing proportion of women presenting with a first episode of acute PID at general practice have received care elsewhere.19 In addition, data from GUM clinics have shown an increase in the number of PID diagnoses in this setting.20 The 48-hour waiting time target for GUM clinics introduced in 200421 may have increased the proportion of cases presenting at GUM clinics. However, as the majority of PID is managed in general practice,12 this is unlikely to fully explain the decline over time.
Changes in the coding of PID by GPs over time could have influenced the decline in diagnoses rates. It is possible that the increase in possible PID diagnoses while definite/probable diagnoses decreased reflects a shift in coding practice toward less specific diagnoses, perhaps as GPs have become more aware of the diagnostic uncertainties. In contrast, it could also be postulated that diagnoses of PID may be more readily made in women with a history of chlamydia diagnosis, and may therefore have become more likely (for the same symptoms) over this period as more women have been tested for chlamydia. Such a “prior-suspicion” effect could be acting to negate, to some extent, a trend of decreasing definite/probable PID incidence. Although there is no evidence to suggest any drastic changes in PID coding practice over the study period, we cannot rule out such changes as a potential contributor to the observed decline in definite/probable PID diagnoses.
Although chlamydia is a common cause of PID, other infections are also associated with the disease, e.g., Neisseria gonorrhoeae.22 Changes in the frequency of these infections, and of their effective treatment, may have contributed to the decline in PID between 2000 and 2008. Rates of gonorrhoea diagnoses among women attending GUM clinics have declined in recent years although this may not reflect population prevalence of untreated infection. However, gonorrhoea has been relatively rare throughout this period.20 Another nonchlamydial factor known to be associated with an increased risk of PID is the insertion of an intrauterine device.23 Use of this method of contraception has remained low in the United Kingdom, with no significant decline, during the study period.24 It seems unlikely that either or both of these other causes of PID could have greatly influenced the observed decline in PID.
A potentially more relevant change over the last decade, is the substantial increases in chlamydia testing and diagnosis rates among women in England—in GUM clinics,20 general practice,10 and more recently, at a broad range of venues participating in the National Chlamydia Screening Programme for England (which was introduced in 2003, with all areas of the country participating by April 2008).8,25 More testing, diagnosis, and treatment of chlamydial infection could have contributed to the declining rates of PID during the study period. It is noteworthy that sharper declines occurred among younger women who are more likely to have been tested and treated for chlamydial infection in more recent years.8,25 A declining rate of PID has also been reported in other developed countries conducting widespread chlamydia testing.26–29 However, population-level surveillance data need to be interpreted with caution given that other factors are also likely to have influenced the observed declines. In England, although overall coverage of chlamydia testing in those aged under 25 years has increased, it remained relatively low during the study period.8,25 It was, therefore, not appropriate to conduct analyses of trends in PID rates by chlamydia testing activity at the local level in this study. However, future analyses should include investigation of regional and local associations between trends in chlamydia testing activity and PID diagnoses, and comparisons of trends in PID among those age groups targeted for chlamydia screening in England (those <25 years) and those not. In addition, comparisons of trends in PID rates pre-, during, and post- the roll out of chlamydia screening (of which our study provides the first 2 elements) will be of value. Further studies will of course be needed to explore the causality of any associations observed between chlamydia testing activity and trends in PID, for instance, cohort studies linking information on chlamydia testing and future PID at the individual level. The GPRD database could provide one source of data for such studies.
An estimated 24% of young women (15–24 years old) in England were screened for chlamydia during April 2008 to March 2009.25 There is evidence from randomized controlled trials in the United States30 and Denmark,31 and the recent Prevention of Pelvic Infection trial in the United Kingdom32 that screening for genital chlamydial infection can reduce the risk of PID (though the reduction was nonsignificant in the UK trial). If the screening program in England is effective, we would expect ongoing increases in the detection and treatment of chlamydial infection to result in greater declines in PID in years to come. We propose to use the methods and case definitions described here for ongoing monitoring of PID diagnoses, as one component of the evaluation of the effect of chlamydia screening in England.
1. Sweet RL. Microbiology. In: Sweet RL, Wiesenfeld HC, eds. Pelvic Inflammatory Disease. London, United Kingdom: Taylor and Francis, 2006:19.
2. Paavonen J, Weström L, Eschenbach D. Pelvic inflammatory disease. In: Holmes KK, Starling PF, Stamm WE, et al, eds. Sexually Transmitted Diseases. New York, NY: McGraw-Hill, 2008:1017.
3. Simms I, Stephenson JM. Pelvic inflammatory disease epidemiology: What do we know and what do we need to know? Sex Transm Infect 2000; 76:80–87.
4. Joesoef MR, Westrom L, Reynolds G, et al. Recurrence of ectopic pregnancy: The role of salpingitis. Am J Obstet Gynecol 1991; 165:46–50.
5. Westrom L, Joesoef R, Reynolds G, et al. Pelvic inflammatory disease and fertility. A cohort study of 1844 women with laparoscopically verified disease and 657 control women with normal laparoscopic results. Sex Transm Dis 1992; 19:185–192.
7. Centers for Disease Control and Prevention (CDC). Chlamydia screening among sexually active young female enrollees of health plans—United States, 2000–2007. MMWR Morb Mortal Wkly Rep 2009; 58:362–365.
9. Medicines and Healthcare Regulatory Agency. General Practice Research Database website. Available at: http://www.gprd.com/home/
. Accessed October 2009.
10. Hughes G, Williams T, Simms I, et al. Use of a primary care database to determine trends in genital chlamydia testing, diagnostic episodes and management in UK general practice, 1990–2004. Sex Transm Infect 2007; 83:310–313.
11. Cassell JA, Mercer CH, Fenton KA, et al. A comparison of the population diagnosed with chlamydia in primary care with that diagnosed in sexual health clinics: Implications for a national screening programme. Public Health 2006; 120:984–988.
12. Simms I, Rogers P, Charlett A. The rate of diagnosis and demography of pelvic inflammatory disease in general practice: England and Wales. Int J STD AIDS 1999; 10:448–451.
13. Simms I, Fleming DM, Lowndes CM, et al. Surveillance of sexually transmitted diseases in general practice: A description of trends in the royal college of general practitioners weekly returns service between 1994 and 2001. Int J STD AIDS 2006; 17:693–698.
14. Simms I, Warburton F, Weström L. Diagnosis of pelvic inflammatory disease: Time for a rethink. Sex Transm Infect 2003; 79:491–494.
15. Maleckiene L, Kajenas S, Nadisauskiene RJ, et al. Comparison of clinical and laparoscopic diagnoses of pelvic inflammatory disease. Int J Gynaecol Obstet 2009; 104:74–75.
16. Ratelle S, Yokoe D, Blejan C, et al. Predictive value of clinical diagnostic codes for the CDC case definition of pelvic inflammatory (PID)—Implications for surveillance. Sex Transm Dis 2003; 30:866–870.
17. Latthe P, Latthe M, Say L, et al. WHO systematic review of prevalence of chronic pelvic pain: A neglected reproductive health morbidity. BMC Public Health 2006; 6:177.
19. Nicholson A, Rait G, Murray-Thomas T, et al. The incidence and management of pelvic inflammatory disease in UK primary care. Results from a large UK primary care database cohort. Br J Gen Pract In press.
22. Bevan CD, Johal BJ, Mumtaz G, et al. Clinical, laparoscopic and microbiological findings in acute salpingitis: Report on a United Kingdom cohort. Br J Obstet Gynaecol 1995; 102:407–414.
23. Farley TM, Rosenberg MJ, Rowe PJ, et al. Intrauterine devices and pelvic inflammatory disease: An international perspective. Lancet 1992; 339:785–788.
24. Lader D, Hopkins G. Omnibus survey report no. 37. Contraception and Sexual Health 2007/08. A report on research using the National Statistics Omnibus Survey produced on behalf of the NHS Information Centre for Health and Social Care. 2008. Available at: http://www.statistics.gov.uk/downloads/theme_health/contra2007–8.pdf
. Accessed October 2009.
26. Kamwendo F, Forslin L, Bodin L, et al. Decreasing incidences of gonorrhea- and chlamydia-associated acute pelvic inflammatory disease. A 25-year study from an urban area of central Sweden. Sex Transm Dis 1996; 23:384–391.
27. Sutton MY, Sternberg M, Zaidi A, et al. Trends in pelvic inflammatory disease hospital discharges and ambulatory visits, United States, 1985–2001. Sex Transm Dis 2005; 32:778–784.
28. Rekart M, Gilbert M, Chang M, et al. Documenting the success of chlamydia control in Br Columbia (BC). Poster presented at: ISSTDR Spring meeting; June 28–July 1, 2009; London, United Kingdom. P4.64.
29. Bohm MK, Newman L, Satterwhite CL, et al. Pelvic inflammatory disease among privately insured women, United States, 2001–2005. Sex Transm Dis 2010; 37:131–136.
30. Scholes D, Stergachis A, Heidrich FE, et al. Prevention of pelvic inflammatory disease by screening for cervical chlamydial infection. N Engl J Med 1996; 334:1362–1366.
31. Ostergaard L, Andersen B, Møller JK, et al. Home sampling versus conventional swab sampling for screening of Chlamydia trachomatis
in women: A cluster-randomized 1-year follow-up study. Clin Infect Dis 2000; 31:951–957.
32. Oakeshott P, Kerry S, Aghaizu A, et al. Randomised controlled trial of screening for Chlamydia trachomatis
to prevent pelvic inflammatory disease: The POPI (prevention of pelvic infection) trial. BMJ 2010; 340:c1642.