This study was conducted using out-patient data obtained from the Hospital of the University of Pennsylvania (called hereafter Penn). Penn contains data from the Philadelphia Metropolitan Area. This includes outpatient data from Southern New Jersey, Philadelphia, parts of Delaware and in the Pennsylvania suburbs of Philadelphia. These data are all structured data collected during routine clinical care between the period of 2006 and 2017. We used women only to demonstrate the effect of visit type on disease associations and also to compare pain diagnoses across cohorts (Fig. 1). Our ‘all clinics’ as shown in Fig. 1 represent all clinics at UPenn where our cohort of women were treated. This includes clinics for cardiology, rheumatology, immunology to name a few. We only investigate women in our analyses because we compare those being treated Ob/Gyn visits versus cancer visits and other visits. Since Ob/Gyn visits generally are made by females it would not make sense to compare against males. In addition many diagnoses, including pain, vary by sex. The demographics of women included in this study are found in Table 1. We included women with an Ob/Gyn visit and outpatient diagnosis code information, women with a cancer visit and outpatient diagnosis code information and all women with diagnoses found in the EHR. The average age for women visiting a cancer clinic was 56 years while those visiting Ob/Gyn clinics were younger (41 years). Because women have multiple visits that can span over several years, we calculated each woman’s average age across their record and then we calculated the overall average and standard deviation shown in Table 1. For defining diagnoses, we used the International Classification of Diseases (ICD) version 9 (ICD-9) and version 10 (ICD-10). The ICD-9 and ICD-10 codes are used primarily for billing purposes to describe the conditions, symptoms, illnesses and diseases for a given patient. We collectively term ICD-9 and ICD-10 diagnoses as conditions in our study.
Importantly, we did not require women have a specific diagnosis of cancer, only that they visited a cancer clinic. Likewise, we did not require that women were pregnant or had a diagnosis of a pregnancy. We only required that they had visited an Ob/Gyn clinic. Women visit Ob/Gyn clinics during all stages of their life, this includes from menarche, childbearing years, during menopause and even after menopause). A total of 1,048 departments and clinics consist within the Penn outpatient health system. Of these 89 are Ob/Gyn/reproductive endocrinology and family planning departments/clinics (8.49%) and only 104 are cancer or oncology departments/clinics (9.92%).
Visit-PheWAS or VisitWAS
We performed a PheWAS using visit type or VisitWAS. This resulted in many disease associations that were correlated with either Ob/Gyn or cancer visits. Overall, there were a total of 7,186 conditions coded for in our dataset. Conditions include diseases (e.g., breast cancer), symptoms (e.g., lump in breast), and infections (e.g., flu) provided that there is a distinct ICD-9 or ICD-10 for that condition. In our VisitWAS study, we performed three different comparisons: 1) those treated at an Ob/Gyn clinic versus any visit (i.e., the entire population), 2) those treated at a cancer clinic versus any visit (i.e., the entire population) and 3) those treated at an Ob/Gyn clinic versus those treated at a cancer clinic. Because patients visiting the Ob/Gyn and those visiting a cancer clinic often visit frequently it made sense to compare these two groups to each other to highlight conditions that were more strongly associated with one group versus the other. We found that many conditions were associated with visit type after adjusting for multiple testing using Bonferroni correction (Additional file 1: Table S1). Out of 7,186 conditions, we found 2,150 significantly associated with an Ob/Gyn visit or 29.92% of all conditions. For cancer visits, we found that 33.58% of conditions were associated (2413/7186). These counts are based on a 0.05 significance threshold that was adjusted for the 7,186 tests we conducted. We also computed the proportion of association conditions that were pain-related along with the proportion of pain conditions associated (out of a total of 129 possible pain conditions).
The Manhattan plots for the association between diseases and visit type are shown in Fig. 2. The Manhattan plots only show the associations for ICD-9 disease categories. We excluded ICD-10 and V codes from the plots. Many of the associations are as expected. Neoplasm diagnoses are strongly associated with cancer visits (lower left hand graph in Fig. 2). Pregnancy, genitourinary and symptom diagnoses are strongly associated with Ob/Gyn visits. Fig. 3 shows the log of the Odds Ratio for the association of diseases to ob./gyn visits vs. cancer visits, which clearly illustrates diseases associated with ob./gyn. Visits (log (OR) > 0) and those associated with cancer visits (log (OR) < 0). Table 2 shows the number of significant associations by ICD-9 disease category broken down by visit type after adjusting for multiple comparisons.
Pain depends on visit type
Overall 129 pain conditions were found in our dataset. We looked at the proportion of associated conditions that were pain related and found that 43 or 2.00% (43/2150) of unique conditions associated with Ob/Gyn visits were pain related (Additional file 1: Table S2). We found that more pain conditions were associated with individuals who visited cancer clinics with 76 or 3.15% of all associations being pain related. For Ob/Gyn visits, we found that 33.33% of pain conditions were associated with Ob/Gyn visits versus 58.91% of pain conditions that were associated with cancer visits. Many of these pain conditions were general and non-specific. When we compared Ob/Gyn vs cancer visits, we found 43 significant pain associations. Only one of these pain conditions was positively associated with Ob/Gyn visits while the remaining 42 pain conditions were associated with cancer visits (or negatively associated with Ob/Gyn visits) as shown in Additional file 1: Table S2. The type of pain unique to Ob/Gyn visits was ‘Pelvis and Perineal pain’ (OR = 2.086). The pain diagnoses that were most associated with cancer visits, i.e., most negatively associated with Ob/Gyn visits (OR < < 1), include ‘neoplasm related pain acute and chronic’ (OR = 0.108 and OR = 0.099 depending slightly on ICD-9 or ICD-10). Also Acute post-thoracotomy pain was heavily associated with cancer visits (OR = 0.243) and Other chronic postoperative pain (OR = 0.359). ‘Chronic pain syndrome’ was also heavily associated with cancer visits with OR = 0.47. We kept the ICD-9 and ICD-10 diagnoses separate because many of the codes in these terminologies differ in terms of their granularity  and granularity can effect the coverage of the code set resulting in mapping difficulties [15,16,17]. Importantly, codes that were exact matches between the two terminologies were both associated and the associations are similar in direction and size.