JMIR Publications

ABSTRACT

Background:

Polycystic ovary syndrome (PCOS) is an endocrine disrupting disorder affecting about 10 percent of reproductive-aged women. In North America and Europe, PCOS diagnosis may be delayed several years and may require multiple doctors resulting in lost time for risk-reducing interventions. Menstrual tracking applications are one potential tool to alert women of their risk for PCOS while also prompting evaluation from a medical professional.

Objective:

The objective of this study was to develop the Irregular Cycles Feature (ICF), an adaptive questionnaire, on the mobile application (app) Clue®, to generate a probability of a virtual test subject’s risk for PCOS. The secondary objective was to evaluate the accuracy of the ICF by comparing the probability of risk generated by the app to a probability generated by a board-certified reproductive endocrinology/infertility physician-scientist (physician).

Methods:

A literature review was conducted to generate a list of signs and symptoms of PCOS, termed variables. These include, but are not limited to hirsutism, acne, and alopecia. Variables were assigned a probability and built into a Bayesian network. Questions were created based on the variables. Virtual test subjects were identified using self-reported menstrual cycles and answers to ICF questions. Upon completion of the questionnaire, a Result Screen summarizing the probability of having PCOS is displayed. A Doctor’s Report containing information regarding menstrual cycles and medical history is also generated. Both documents provide information about PCOS and data that may facilitate diagnosis by a medical professional. To assess the accuracy of the ICF, virtual test subjects were assigned probabilities by a) the ICF and b) the physician, who served as the gold standard. The ICF recommends individuals with a score greater than or equal to 25% to follow-up with a physician. Differences between the network and physician probability scores were evaluated using a t-test and a Pearson correlation coefficient. A second iteration of the ICF was done to assess the ICF’s probability capabilities.

Results:

The ICF’s first iteration produced one false positive compared to the physician score and had an absolute mean difference of 15.5% (SD= 15.1%) amongst virtual test subjects. The second iteration had two false positives compared to the physician score and had an absolute mean difference of 18.8% (SD = 13.6%). The ICF overpredicted virtual test subjects’ risk of PCOS compared to the physician. However, a strong positive significant correlation existed between the ICF and the physician score (Pearson correlation coefficient= 0.69; p < 0.01). The second iteration performed worse with a Pearson correlation coefficient of 0.54; p > 0.01).

Conclusions:

The first iteration ICF outperformed the second, and better predicted the probability of PCOS. The ICF can potentially be used as a screening tool to prompt high-risk subjects to seek evaluation by a medical professional.