Article Text

Download PDFPDF

Missing pelvic inflammatory disease? Substantial differences in the rate at which doctors diagnose PID
  1. A Doxanakis1,2,
  2. R D Hayes2,
  3. M Y Chen1,2,
  4. L C Gurrin2,
  5. J Hocking2,
  6. C S Bradshaw1,3,
  7. H Williams1,2,
  8. C K Fairley1,2
  1. 1
    Melbourne Sexual Health Centre, Melbourne, Victoria, Australia
  2. 2
    School of Population Health, The University of Melbourne, Victoria, Australia
  3. 3
    Department of Epidemiology and Preventive Medicine, Monash University, Prahran, Victoria, Australia
  1. Christopher K Fairley, University of Melbourne, Director Melbourne Sexual Health Centre, 580 Swanston Street, Carlton Victoria 3053, Australia; cfairley{at}mshc.org.au

Abstract

Objectives: The clinical diagnosis of pelvic inflammatory disease (PID) is subjective. Our aim was to determine if the pattern of diagnosis of PID among experienced clinicians varied compared with the diagnosis of genital warts.

Methods: We conducted a retrospective study of 325 PID diagnoses made by experienced clinicians at Melbourne Sexual Health Centre, Australia (2002–2006), where doctors saw 21 785 unselected female patients in a walk-in service. We compared the proportion of female patients diagnosed as having PID and genital warts between doctors and then compared doctors above (high diagnosing) and below (low diagnosing) the mean rate of PID diagnosis.

Results: There were significant and clinically important differences in the proportion of women diagnosed with having PID (0–5.7%) across 23 doctors investigated. Estimated standard deviation in the frequency of PID diagnosis (logit scale) was 1.26 (95% CI 0.81 to 1.95)—approximately four times greater than for warts. Patients seen by high (n = 4673) and low (n = 16 787) diagnosing doctors had similar epidemiological risk profiles suggesting true distribution of PID cases across doctors was similar (p>0.13). Women diagnosed with having PID by high diagnosing doctors compared with low diagnosing doctors were younger (odds ratio 1.7; 95% CI 1.1 to 2.8, p = 0.013) but otherwise had similar epidemiological and clinical features.

Conclusions: Differences in diagnostic rates for PID between doctors are substantial and may be because of PID cases being missed by some doctors.

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Pelvic inflammatory disease (PID) is a substantial cause of reproductive morbidity globally. Left untreated, or if treatment is delayed, PID can result in serious reproductive complications, including infertility, chronic pelvic pain and an increased risk of ectopic pregnancy. If we are to prevent such complications it is important that we ensure that all women with PID receive early treatment.1

Diagnosis of PID on clinical grounds is often subjective and hence there is the potential for considerable differences between clinicians in making a diagnosis.2 3 Women with PID present with a wide range of clinical symptoms and signs that range from subclinical infection to acute, severe PID. Definitions and guidelines for the diagnosis of PID have been proposed by the Centers for Disease Control and Prevention (CDC);4 however, there is some evidence that the CDC’s minimal criteria for the diagnosis of PID have relatively low sensitivity and specificity.5 6 The “gold standard” of diagnosis of PID is laparoscopy, but even this is not 100% sensitive and its use in outpatient settings is either impractical or not indicated.7

Although difficulties with the diagnosis of PID may lead to differences in the rate of diagnosis made between clinicians, no previous studies have examined this issue. Our aim was to determine how consistent the diagnosis of PID was among a group of sexual health clinicians who commonly see women at risk for PID and to compare this with differences in the diagnosis of genital warts—a condition where clinical diagnosis is less subjective.

METHODS

The Melbourne Sexual Health Centre (MSHC) is the only public sexual health clinic serving the city of Melbourne, Australia. We conducted a retrospective study investigating the proportion of women diagnosed with having PID and genital warts attending MSHC between 1 July 2002 and 30 June 2006. We also assessed the risk profile of women seen by different doctors during the study period. The MSHC provides a walk-in triage system that prioritises access to the centre for patients who belong to risk groups or who have symptoms. Patients cannot make appointments except for review after an initial consultation, and these are only available for up to 4 weeks after the initial consultation.8 Referrals are not required and there are no fees associated with consultations, investigations and treatments.

Doctors at the MSHC take patients in the order in which they are triaged in. After a doctor has assessed a patient, they enter a diagnosis from a predetermined list that includes PID and warts. For new patients and returning patients not seen within 3 months, epidemiological data are also collected and entered electronically, including the number and gender of sexual partners in the last 3 and 12 months, condom use, current sex work and history of injecting drug use. Diagnoses and epidemiological data are entered into the clinic’s computerised database.

In the first analysis, cases were defined as female patients presenting to the MSHC between 1 July 2002 and 30 June 2006 who were diagnosed with having PID. Controls were all the other female patients presenting during the same period who were not diagnosed with having PID. Demographical and epidemiological data were then compared. Cases or controls where epidemiological data were missing (that is, those returning within less than 3 months) were excluded from this analysis.

Clinical record audit

The clinical files of all women diagnosed with having PID were reviewed if they were seen by permanent senior medical staff of the centre. Junior trainee doctors on short-term rotations were excluded. Women seen by nurses were also excluded from the analysis. Clinical data were extracted only on the day PID was diagnosed. The data included epidemiological risk factors, contraceptive use (for example, oral contraceptives, intrauterine device), past history of PID, past gonorrhoea or chlamydia, symptoms of lower abdominal pain, vaginal discharge, dysuria, deep dyspareunia, fever, inter-menstrual bleeding and post-coital bleeding.

We then compared the rate of PID diagnoses made between different doctors. We also compared the rate of diagnosis of PID with the rate of diagnosis of genital warts—a control condition about which there is little controversy surrounding the criteria for diagnosis. Doctors included in this analysis were classified into high and low PID diagnosing groups on the basis of their mean rate of PID diagnosis. Low diagnosing doctors were defined as those who had diagnostic rates for PID less than or equal to the mean rate among all doctors, whereas high diagnosing doctors had rates of diagnosis above the mean. The characteristics of women diagnosed by high and low diagnosing doctors were then compared. This was a clinical quality assurance audit and under Australian guidelines did not require formal ethics approval.

Statistical analysis

Odd ratios (OR), 95% CI and p values were calculated for categorical, ordinal and numerical data as appropriate. The diagnoses of PID and warts were analysed using a random effects logistic regression model,9 which quantifies explicitly the variation in the frequency of diagnosis across doctors and by calculating the intra-cluster correlation coefficient (ICC)10 to determine the extent to which diagnoses cluster within doctors.

RESULTS

Over the 4-year study period, there were a total of 37 127 female consultations at MSHC. This included 36 741 consultations for women without a diagnosis of having PID and 386 consultations where a diagnosis of PID was made. Twenty of these cases of PID were in women with repeated episodes of PID, leaving 366 new PID diagnoses. Of these 366 cases, 41 were seen by doctors in training, leaving 325 cases in the analysis. There were 21 460 controls remaining after exclusion of those seen by nurses and junior doctors.

Compared with controls, women with PID were younger (OR 0.6; 95% CI 0.5 to 0.8, p<0.001), more likely to have had three or more partners in the last 3 months (OR 1.7; 95% CI 1.1 to 2.5, p = 0.007), more likely to have had four or more partners in the last year (OR 1.8; 95% CI 1.4 to 2.4, p<0.001), more likely to have used condoms inconsistently during the last 3 months (OR 2.2; 95% CI 1.5 to 3.5, p<0.001), more likely to have used condoms inconsistently during the last year (OR 1.8; 95% CI 1.2 to 2.7, p = 0.004) and less likely to be a sex worker (OR 0.4; 95% CI 0.2 to 0.5, p<0.001). No significant difference was detected with regard to injecting drug use (OR 1.3; 95% CI 0.8 to 2.0, p = 0.28).

Table 1 lists the number of PID and wart diagnoses and the total number of female consultations for each of the 23 permanent non-trainee doctors who were included in this analysis. The percentage of consultations that resulted in a PID diagnosis ranged from 0–5.7% across doctors with a mean diagnostic rate of 1.7%. The corresponding range in the frequency of wart diagnoses was 1.3–6.9% with a mean of 5% of consultations resulting in diagnoses.

Table 1 Proportion of female clients diagnosed with pelvic inflammatory disease (PID) or warts by doctors

The estimated standard deviation in frequency of PID diagnosis (on the logit scale) was 1.26 (95% CI 0.81 to 1.95)—approximately four times greater than the corresponding estimate for warts of 0.31 (95% CI 0.18 to 0.49). The ICC was less than 2% for both PID and warts. This suggests that the proportion of variation in the frequency of diagnoses that can be attributed to differences in doctors’ diagnosis frequencies is low for both conditions, although the ICC is approximately five times higher for PID than warts. Even if we excluded data from the four doctors whose frequencies of PID diagnosis were highest (5.7%, 5.1%, 5.0% and 4.8%, the next lowest is 2.0%), the estimated standard deviation for PID is 0.99 (95% CI 0.54 to 1.71). This is still three times greater than the standard deviation estimate for warts. In subsequent analyses, high diagnosing doctors were defined as those who diagnosed PID in more than 1.7% of female consultations (the mean rate of diagnosis among all the doctors at the service). According to this definition, six doctors were arbitrarily classified as being high PID diagnosing doctors.

To determine whether there was any apparent bias in the selection of women by low and high diagnosing doctors, the 4673 controls who presented to high PID diagnosing doctors were compared with the 16 787 controls who presented to low PID diagnosing doctors (table 2). No differences were detected between controls presenting to high and low PID diagnosing doctors with respect to any of the variables investigated (p>0.13): these included age, male sexual partners in the last 3 months, male sexual partners in the last year, condom use in the last 3 months, condom use in the last 6 months, sex worker status or injecting drug use.

Table 2 Comparison of the characteristics of women without pelvic inflammatory disease (PID) (controls) seen by high and low PID diagnosing doctors

Epidemiological characteristics, symptoms and signs, and test results of women diagnosed with having PID by high diagnosing doctors (n = 160) were compared with those of women diagnosed with having PID by low diagnosing doctors (n = 165) (table 3). Patients diagnosed by high diagnosing doctors were significantly more likely to be younger (OR 1.7; 95% CI 1.1 to 2.8); however, no significant differences were detected with respect to the remaining patient characteristics investigated. In addition, no significant differences were detected for any of the patient symptoms and signs (p>0.09). There was also no evidence that investigation results were different for cervical polymorphonuclear cell count, vaginal polymorphonuclear cell count, chlamydia test result and diagnosis of bacterial vaginosis (p>0.28). Further data (not shown in the tables) indicated that there were no significant differences with respect to injecting drug use, inter-uterine device use, oral contraceptive use or past gonorrhoea. Using the CDC definition for PID,6 there was some evidence high diagnosing doctors were less likely than low diagnosing doctors to fulfil this criteria when making a diagnosis of PID (52% vs 62%; OR 1.5; 95% CI 0.97 to 2.5).

Table 3 Pelvic inflammatory disease (PID) case group: comparison of patients seen by high diagnosing doctors and low diagnosing doctors

There was no statistical difference between high diagnosing and low diagnosing doctors with respect to their gender (50% female vs 24% male, p = 0.23) or having completion of specialist training in sexual health (67% vs 53%, p = 0.56). However, this analysis had limited statistical power. To further verify that diagnoses of PID entered onto the database reflected a clinician’s intention to treat the women for PID, we also obtained the pharmacy records of prescriptions for 2 weeks of doxycycline, which is almost exclusively used for the treatment of PID at MSHC. The number of scripts for this given by each of the doctors was almost exactly the same as the PID diagnoses for each doctor.

DISCUSSION

In this study of women attending a sexual health clinic, there was a large and clinically important difference in the rate of PID diagnoses between experienced clinicians. By contrast, a much smaller difference in diagnostic rates occurred for genital warts. The difference in PID diagnoses occurred despite all women having a similar epidemiological risk profile, suggesting that the difference in the rate of PID diagnoses is not likely to be due to a difference in the true prevalence of PID among the women seen by different doctors.

The implications of these findings are that either large numbers of women are having PID diagnosis missed or that PID is being over diagnosed. If high diagnosing doctors were over diagnosing PID then we would expect a smaller proportion of the cases diagnosed by them to display the characteristic epidemiological and clinical features of PID since many of these women would not truly have PID. Instead, we found no differences (with the exception of age) in the characteristics of women with PID when those diagnosed by high and low diagnosing doctors were compared. These data are consistent with high diagnosing doctors detecting greater numbers of real PID cases compared with low diagnosing doctors; therefore, it is likely that, overall, many cases of PID are being missed. Although we were unable to show why there were these differences, we would postulate that some doctors had a higher “threshold” for diagnosing PID. That is, they required a higher number of clinical criteria or more severe clinical features to be present before making the diagnosis.

Our findings are predicated on a number of important assumptions. The first is that women seen by different doctors have a similar chance of having PID. This is supported by the fact that women without a diagnosis of PID (table 2) seen by high and low diagnosing doctors had the same epidemiological risk profile with all point estimates of odds ratios being between 0.9 and 1.1. A sample size of many thousands in each group ensured substantial statistical power. Moreover, other studies we have undertaken have shown that the walk-in triage system allocates clients to doctors in such a way that they are epidemiologically similar.11 The second important issue in interpreting our findings is ensuring that the coding of PID diagnoses was accurate. The treatment recommendations at MSHC for PID include a 2 week course of doxycycline. The number of prescriptions for this regimen from each of the doctors closely matched the number of PID diagnoses made during the study period.

However, this study does have a number of weaknesses. First, we cannot verify the accuracy of PID diagnoses that were made. Diagnoses were made on clinical grounds, without laparoscopic or histological confirmation, reflecting common outpatient practice in our setting.12 13 That said, the risk factors associated with PID among women in this study were highly consistent with risks that have been previously identified for PID in other studies.14 Furthermore, the PID cases that were diagnosed in this study were those that were clinically overt and do not include women with subclinical PID, defined as histologically proven endometritis in the absence of symptoms or signs of PID, a condition that has been implicated in clinically significant reproductive sequelae such as fallopian tube damage.1517 As this study was retrospective, it was not possible to ascertain the severity of clinical symptoms and signs.

Given how subjective the diagnosis of PID can be and how deleterious its consequences can be for women, it has been recommended that in practice the threshold for diagnosing and treating PID should be low. Our study shows that in our setting, where experienced clinicians are managing women at high risk for this condition, this is not the case among many doctors.

Key messages

  • There was a substantial and clinically important difference in the rate of diagnosis of pelvic inflammatory disease (PID) made by different doctors.

  • This difference was much less for the diagnosis of genital warts.

  • This may mean that a large number of PID diagnoses are being missed by some doctors.

Acknowledgments

We would like to thank Jun Kit Sze for the data extraction necessary for this study.

REFERENCES

Footnotes

  • Funding: C S Bradshaw holds a National Medical & Research Council Research Scholarship. No external sources of funding were used to support this project.

  • Competing interests: None.

  • Ethics approval: This was a clinical quality assurance audit and under Australian guidelines did not require formal ethics approval.

  • Contributors: All authors contributed to the design and planning of this study. AD was responsible for the chart reviews. AD, MYC and CKF were responsible for the database analysis. JH and LG oversaw the statistical analysis. All authors contributed to the writing, editing and approval of the manuscript.

Linked Articles

  • Brief encounters
    Helen Ward Rob Miller