Statistical Process Control In IVF
Cloud-based electronic medical record systems permit the analysis of large amounts of IVF data on a granular basis. Preliminary investigations suggest that Statistical Process Control (SPC) analysis of variation within an IVF center may be predictive of quality within ART unit and that high process variation may be associated with poorer IVF outcomes.
In a preliminary study SPC analysis was applied to two ART units:
• Clinic A: solo practitioner performing ~100 cycles per year. ART treatments performed in a high volume national-class IVF laboratory with advanced air handling systems.
• Clinic B: solo practitioner performing ~ 80 cycles per year. ART treatments performed by a consulting embryologist in a facility lacking specialized air handling technology.
Mean ages for patients treated in the two Clinics were not different.
Review of SART pregnancy rates for women aged < 35 demonstrates a marked difference in outcome:
In this study we performed statistical process control analysis for metrics of IVF clinical and laboratory quality, including number of oocytes harvested, percent fertilization, and embryo quality.
To reduce the outcome of an individual IVF procedure to a simple metrics for analysis, each embryo was assigned an "embryo score" by multiplying the number of cells by the points assigned for embryo quality
(A = 6 points, B = 3 points, C = 0).
The "mean embryo score" (MES) for each procedure was calculated by dividing the sum of the individual embryo scores by the number of 2 PN embryos.
Successful cases had MES in the 30s and 40s, with a majority of embryos in the upper right portion of the grid:
Poorer cases had both lower MES with embryos clustering in the bottom left portion of the grid:
Control charting all patients for Clinic A demonstrated characteristics of a stable process. MES scores were tightly distributed around a high mean. Many of the outlying points were the result of patient characteristics such as poor stimulation or impaired ovarian reserve (Special Cause Variation not attributable to laboratory performance):
Preliminary control charting demonstrated significant variation due to patient characteristics. Upper and Lower Control Limits (mean +/- 3 SD) exceeded the range of values for MES (0 to 60). Hence Control Limits of mean +/- 1 SD are shown for illustration purposes.
In contrast, control charting all patients for Clinic B demonstrates significantly more process variation. Average MES was lower than Clinic A with a higher standard deviation. Areas of widely fluctuating values and clusters of values above or below the mean are consistent with a much less stable process than Clinic A.
Creating "Idealized Patients" to Adjust SPC for Patient Attributes
SPC for Clinic A demonstrated a higher average MES with a lower standard deviation as compared to the unselected patient population. In contrast, even with a selected group of optimal quality patients, Clinic B demonstrated significantly more process variation:
The common retort of IVF units having lower pregnancy success rates is that
"my patients are harder than everyone else's". This may be true. Or not.
To address this issue we performed SPC for a subset of high quality patients - women aged <36 that produced 8 or more eggs.
Using SPC to Identify Critical Process Steps
If SPC can differentiate between high performing and lower performing ART units on the basis process variation, then the question is whether SPC can identify specific steps of the ART process in need of remediation.
To explore this question control charts were constructed using "idealized patients" for the steps of oocyte aspiration, fertilization, and embryo culture to day three.
Control charting for numbers of oocytes harvested revealed similar patterns between the two Clinics:
Similarly, percent fertilization for "idealized patients" were relatively similar, although there appears to be a trend toward increased variation in Clinic B.
In contrast, variation in day three MES, as shown above, varies markedly between the two laboratories:
This suggests that the principle difference in performance between the two Clinics lies in the step of embryo culture. Clearly, the next step in process management for Clinic B is to identify Special Causes of Variation in the areas of laboratory environmental control and technician skill in an effort to improve process quality.
These preliminary studies offer a glimpse into the utility of SPC to assess process variability as a means of objectively quantifying IVF laboratory quality. Further investigation is needed to determine whether SPC will be useful to detect quality issues at specific aspects of IVF medical and laboratory process. SPC may have use in the process of training and assessing clinical and laboratory skills and competence.
© Copyright 2011. Artisan Medical Solutions LLC. All Rights Reserved.