First year of radiology residency
200 days worth of statistics
This Friday I did my 200th shift for my residency in radiology. A little more than a month ago, I started my second year of residency, and during the last 1.5 years (since I work part time) I’ve compiled statistics over various aspects of my education. Both to keep track of my development, but also for fun.
For example, I’ve kept track of the number of each examination I’ve performed in order to identify whether I am lacking in experience with regard to specific examinations or body parts.
I’ve also kept track of the active time I’ve spent on each examination, which allows me to both calculate a measure of efficiency (examinations per hour), but also to correct for the relative difficulty and time it takes to interpret various examinations. It takes much more time to perform an ultrasound or interpret a CT than a regular X ray. I’ve used that to get a weighted efficiency score which is a much better indicator of actual workload. It allows me to compare my efficiency across days since it is independent of the types of examinations I happen to be allocated depending on the position and shift.
I’ve also recorded the number and type of findings in the examinations (both main diagnoses, and secondary findings), as well as things such as distance travelled by bike (we have bikes for our work stations to increase our activity levels). I have:
performed or interpreted 3960 examinations
2302 X ray images
1358 CT scans
277 ultrasounds
14 MRI scans
9 fluoroscopies
5 CT/ultrasound-guided interventions/biopsies
worked actively for 1118 hours, for a total of 3,5 examinations per hour
done 62 on-call evening and night shifts
held 19 radiology rounds for clinicians
tutored medical students 19 times, for lectures, seminars and as clinical supervision
answered the phone 800 times, and prioritised 1192 referrals, while biking 303 km
LEARNING DIAGNOSTIC RADIOLOGY
Over time, by gaining knowledge and getting my eyes used to each examination, I’ve improved my speed and efficiency. The plot below shows how my efficiency has increased across time. As expected, there was a rapid initial improvement, with a slow decrease in the slope of the red curve (best fit line is a third-degree polynomial because I haven’t yet been able to completely regress out the effect of different rotations since I’ve only done two), which indicates my expected skill level.
The black line shows the residuals for each day (distance to red line), which indicates how efficient I was that day, irrespective of my skill level. This allows for comparisons across time, such as between early days when I was a beginner and now.
There is a lot of noise in the data, since I have vastly different tasks depending on which position I have. Sometimes I’m coordinating the inflow of referrals (and doing few, if any examinations myself), and sometimes I’m the only radiologist on call doing tons of examinations.
By keeping track of which positions I have each day, I’ve been able to calculate the difference in mean residual (efficiency) for those positions compared with all other positions. For example, I perform 3.5 more (weighted) examinations per hour when on call, and 1 less when preparing rounds (since I mostly do not interpret all of them myself).
My efficiency is slightly levelling off because it is essentially a measure of my radiological knowledge (it is rising again since I recently started doing fluoroscopies and MRI, and thus am continually learning, but has almost completely levelled of when just looking at on-call shifts). The main limiting factor is that I have to either ask a specialist or search for information myself when confronted with an unknown finding or diagnosis. At the beginning of the residency, one cannot complete a single examination without consulting both colleagues and the literature. Now, after 200 days and 62 on-call shifts, I rarely encounter something in emergency radiology that I have to look up.
TRAINING THE EYEs AND NOT ONLY THE BRAIN
Radiologists consist not only of a brain containing theoretical knowledge, but also of eyes used to identify conditions. The diminishing returns above is due to the limited number of specific diagnoses and examinations. The actual visual differences are vastly more numerous due to different examinations, parts of examinations and the body as well as different diagnoses. This means that the eyes have a much longer learning period than the brain, in terms of becoming an experienced and proficient radiologist. The plots below illustrate this. (Note: the days are mixed because I did on-call shifts at Mölndal hospital while working at Östra Hospital.)
On the left plot (positive findings per examination) one can see that I was fairly stable around 1 for my entire first rotation. This is because I was focused on musculoskeletal radiology and fractures only, and they either had a fracture or they didn’t. Only later, when I started my abdominal rotation did my brain and eyes get used to the large number of secondary findings, and their existence. One can tell that I, for my second musculoskeletal rotation, improved my sensitivity to these findings.
Even though only a potential fracture may be noted in the radiological report, the patient can also have kidney cysts and arthritis in the hip, for example. The exponential increase in my rate of identification of positive findings means I am not stagnating, and illustrates the point I made earlier. This is practically illustrated by a resident and a specialist both knowing what appendicitis looks like, but the specialist being able to identify the inflamed appendix, as well as a small abscess, free air/perforation, and regional lymphadenopathy by scrolling through the CT once, while the resident may have to scroll through more than once for each finding/complication they are looking for. The right plot also takes into account my working speed, superimposing the rate of learning across time on the rate of identification of positive findings.
I’ve made a point in actively trying to identify as many positive findings as possible in each examination I interpret. I believe this is crucial in the education of a radiologist, since it forces you to look at all structures and know their normal appearance. This may seem like a waste of time in the beginning, but I don’t think one can gain the visual experience passively, and it likely contributes to increased speed of interpretation across time, and will serve to increase production speed in my entire future career as a radiologist.
EFFECT OF COVID-19
I have also been able to quantify the negative impact that the Covid-19 pandemic has had on my education. During the first wave in Sweden, people were afraid to visit the emergency departments, and much of the planned healthcare, including radiological examinations, was put on hold. This led to a strong decrease in volume at our clinics.
This plot shows my progression in the number of (weighted) examinations per working hour. Although there is no clear beginning or end to a wave, I’ve shaded the dates during which my hospital (Sahlgrenska University Hospital) increased their emergency preparedness and decreased production. I’ve fitted a curve to the raw data outside those dates (red line) to get a sort of expected development, if it weren’t for the change in policy. The black line shows the moving average for the actual number of weighted examinations. By calculating the area between the curves and multiplying by the hours spent working each day within the shaded area I was able to estimate a net loss of 334 examinations. Given that I recently (during my 201st shift yesterday) completed my 4000th examination, this means that I’m 8% less experienced as a radiologist than I would’ve been without the pandemic.
PART TIME WORK
I usually work clinically three days per week. I often get home after work and read up on some of the diagnoses I’ve identified during the day. If I work several days consecutively, I may not do this (I plan to do it the next day at work, which may not happen depending on the workload that day). To identify the optimal ratio of clinical and non-clinical days, I’ve plotted my efficiency against the rate of clinical days. A value of 2 means that each of the three consecutive clinical days were separated by at least one non-clinical day (1 means a fill-time schedule).
There is a clear inverted U-curve; when working closer to full-time there is less time and energy to study and keep the production high; when working infrequently, one falls out of shape and has time to forget the specific radiological knowledge.
SUMMARY
Those who follow my blog know that I am all about objective quantification as a means of training, learning, and improving. Naturally this extends to my radiological education. The findings I present here are a minuscule part of the data I have and the analyses I’ve performed.
Hopefully, I’ve been able to show how this approach can help guide and optimise, even something as diffuse as a radiological education. By identifying diminishing returns in one area or type of examination, one knows when to move on to the next, and thus optimise longitudinal development. And by keeping track of which examinations one has performed, both regarding body parts and diagnoses, one can identify holes in ones’ experience that need to be filled before becoming a self-sufficient specialist.
In order to simplify the data collection, I’ve created a form that I fill out continually while I work, as well as an excel spreadsheet to which i transfer all the data. Once you get used to it, it takes an average of 5 seconds to add an examination and its findings to the form, and an additional 4-5 minutes to transfer the data to the spreadsheet at the end of the day. You are free to download both the form and (a simplified version of the) spreadsheet to allow for easy data collection. For now, I only have a Swedish version; in case someone wants to use it, let me know and I can translate it to English.