Sanjat Kanjilal
@sanjatkanjilal.bsky.social
Group leader | Dept of Medical Microbiology | Amsterdam University Medical Center @amsterdamumc.bsky.social | research machine learning for dx/tx/prevention of infectious diseases
When you can watch an entire series with a thoroughly unlikeable main character, you know the director and actor have done an amazing job!
Also, the attention to detail in the set pieces somehow manages to capture the beauty of that time and place without empty sentimentality, brilliant
Also, the attention to detail in the set pieces somehow manages to capture the beauty of that time and place without empty sentimentality, brilliant
November 27, 2025 at 12:00 PM
When you can watch an entire series with a thoroughly unlikeable main character, you know the director and actor have done an amazing job!
Also, the attention to detail in the set pieces somehow manages to capture the beauty of that time and place without empty sentimentality, brilliant
Also, the attention to detail in the set pieces somehow manages to capture the beauty of that time and place without empty sentimentality, brilliant
Gave my 1st talk in NL: how to make multicenter data analysis with patient data actually work.
tl;dr: common data models (global interoperability) + LLMs (data harmonization, mapping, future-proofing) + federated learning (privacy)
Let me know what you think!
drive.google.com/file/d/1N3yk...
tl;dr: common data models (global interoperability) + LLMs (data harmonization, mapping, future-proofing) + federated learning (privacy)
Let me know what you think!
drive.google.com/file/d/1N3yk...
November 5, 2025 at 11:06 AM
Gave my 1st talk in NL: how to make multicenter data analysis with patient data actually work.
tl;dr: common data models (global interoperability) + LLMs (data harmonization, mapping, future-proofing) + federated learning (privacy)
Let me know what you think!
drive.google.com/file/d/1N3yk...
tl;dr: common data models (global interoperability) + LLMs (data harmonization, mapping, future-proofing) + federated learning (privacy)
Let me know what you think!
drive.google.com/file/d/1N3yk...
A stroll around the new hood
October 3, 2025 at 12:46 PM
A stroll around the new hood
That's a stage being setup...inside @amsterdamumc.bsky.social...where the hospital's band (yup), play a concert once or twice a year. How about that!
September 9, 2025 at 7:52 AM
That's a stage being setup...inside @amsterdamumc.bsky.social...where the hospital's band (yup), play a concert once or twice a year. How about that!
I gave my final didactic talk to the @mgb-id-fellows.bsky.social fellows yesterday on treatment of MDR GNRs. Really active area of research!
Have a look! I welcome corrections.
drive.google.com/file/d/170xP...
Have a look! I welcome corrections.
drive.google.com/file/d/170xP...
July 17, 2025 at 1:06 PM
I gave my final didactic talk to the @mgb-id-fellows.bsky.social fellows yesterday on treatment of MDR GNRs. Really active area of research!
Have a look! I welcome corrections.
drive.google.com/file/d/170xP...
Have a look! I welcome corrections.
drive.google.com/file/d/170xP...
I feel you dude
June 29, 2025 at 1:51 PM
I feel you dude
Waterbury VT shows up!
June 15, 2025 at 1:31 PM
Waterbury VT shows up!
Interestingly, we noted some inverse relationships (eg model-adjusted CP values were higher in controls than in cases), mostly between organisms that inhabit different niches.
For instance, CP-MSSA ↓ odds of ESBL K. pneumoniae by -7.9%; CP-drug-R PsA and ↓ odds of VSE. faecalis by -10.0%
10/14
For instance, CP-MSSA ↓ odds of ESBL K. pneumoniae by -7.9%; CP-drug-R PsA and ↓ odds of VSE. faecalis by -10.0%
10/14
June 12, 2025 at 11:03 AM
Interestingly, we noted some inverse relationships (eg model-adjusted CP values were higher in controls than in cases), mostly between organisms that inhabit different niches.
For instance, CP-MSSA ↓ odds of ESBL K. pneumoniae by -7.9%; CP-drug-R PsA and ↓ odds of VSE. faecalis by -10.0%
10/14
For instance, CP-MSSA ↓ odds of ESBL K. pneumoniae by -7.9%; CP-drug-R PsA and ↓ odds of VSE. faecalis by -10.0%
10/14
Among the skin flora, CP-MSSA ↑ odds of MSSA by +12.1% & CP-MRSA ↑ odds of MRSA by +6.7%
Among the environmental flora, CP-drug-S-PsA ↑ odds of drug-S-PsA by +10.1% and CP-drug-R-PsA ↑ odds of drug-R-PsA by +28.6%
9/14
Among the environmental flora, CP-drug-S-PsA ↑ odds of drug-S-PsA by +10.1% and CP-drug-R-PsA ↑ odds of drug-R-PsA by +28.6%
9/14
June 12, 2025 at 11:03 AM
Among the skin flora, CP-MSSA ↑ odds of MSSA by +12.1% & CP-MRSA ↑ odds of MRSA by +6.7%
Among the environmental flora, CP-drug-S-PsA ↑ odds of drug-S-PsA by +10.1% and CP-drug-R-PsA ↑ odds of drug-R-PsA by +28.6%
9/14
Among the environmental flora, CP-drug-S-PsA ↑ odds of drug-S-PsA by +10.1% and CP-drug-R-PsA ↑ odds of drug-R-PsA by +28.6%
9/14
Top line results
Nosocomial acquisition occurred 4X more frequently for drug-S than drug-R bugs
For enteric flora, CP-C. diff ↑ odds of C. diff by +32.5%. ESBL-Enterobacterales ↑ odds of ESBL-Kpn by +29.4%
CP-VSE ↑ odds of ESBL E. coli (+5.0%), VSE. faecalis (+11.8%) & VSE. faecium (+8.6%)
8/14
Nosocomial acquisition occurred 4X more frequently for drug-S than drug-R bugs
For enteric flora, CP-C. diff ↑ odds of C. diff by +32.5%. ESBL-Enterobacterales ↑ odds of ESBL-Kpn by +29.4%
CP-VSE ↑ odds of ESBL E. coli (+5.0%), VSE. faecalis (+11.8%) & VSE. faecium (+8.6%)
8/14
June 12, 2025 at 11:03 AM
Top line results
Nosocomial acquisition occurred 4X more frequently for drug-S than drug-R bugs
For enteric flora, CP-C. diff ↑ odds of C. diff by +32.5%. ESBL-Enterobacterales ↑ odds of ESBL-Kpn by +29.4%
CP-VSE ↑ odds of ESBL E. coli (+5.0%), VSE. faecalis (+11.8%) & VSE. faecium (+8.6%)
8/14
Nosocomial acquisition occurred 4X more frequently for drug-S than drug-R bugs
For enteric flora, CP-C. diff ↑ odds of C. diff by +32.5%. ESBL-Enterobacterales ↑ odds of ESBL-Kpn by +29.4%
CP-VSE ↑ odds of ESBL E. coli (+5.0%), VSE. faecalis (+11.8%) & VSE. faecium (+8.6%)
8/14
Our key predictor is colonization pressure (CP), defined as the time-weighted prevalence of a bug among ward co-occupants the day a new patient arrives. We looked back in co-occupants over 6 months in case they had silent carriage & down-weighted events occurring farther back in time
6/14
6/14
June 12, 2025 at 11:03 AM
Our key predictor is colonization pressure (CP), defined as the time-weighted prevalence of a bug among ward co-occupants the day a new patient arrives. We looked back in co-occupants over 6 months in case they had silent carriage & down-weighted events occurring farther back in time
6/14
6/14
We went to great lengths to minimize differences between cases & controls. We matched on a bunch of factors, including 14 different classes of abx; not just whether they received an abx or not, but how many courses. We also made sure they weren't coming in w/infxn and stayed in only 1 room
5/14
5/14
June 12, 2025 at 11:03 AM
We went to great lengths to minimize differences between cases & controls. We matched on a bunch of factors, including 14 different classes of abx; not just whether they received an abx or not, but how many courses. We also made sure they weren't coming in w/infxn and stayed in only 1 room
5/14
5/14
Even if science was more transparent / reproducible, I think we’d be in the same place. This reflects a gradual loss of critical thinking for the public, exacerbated by income inequality and for profit corporations who profit off ignorance. Solutions aren’t easy.
Carl Sagan said it best.
Carl Sagan said it best.
June 11, 2025 at 11:11 AM
Even if science was more transparent / reproducible, I think we’d be in the same place. This reflects a gradual loss of critical thinking for the public, exacerbated by income inequality and for profit corporations who profit off ignorance. Solutions aren’t easy.
Carl Sagan said it best.
Carl Sagan said it best.
Hi world, I have exciting news to share: I will be joining the faculty of the @amsterdamumc.bsky.social this fall!
I am thrilled to start afresh with a fantastic group of colleagues and a highly supportive academic environment.
A new adventure awaits!
I am thrilled to start afresh with a fantastic group of colleagues and a highly supportive academic environment.
A new adventure awaits!
May 20, 2025 at 7:27 PM
Hi world, I have exciting news to share: I will be joining the faculty of the @amsterdamumc.bsky.social this fall!
I am thrilled to start afresh with a fantastic group of colleagues and a highly supportive academic environment.
A new adventure awaits!
I am thrilled to start afresh with a fantastic group of colleagues and a highly supportive academic environment.
A new adventure awaits!
The regime just killed the crown jewel of my med school.
May 16, 2025 at 11:57 AM
The regime just killed the crown jewel of my med school.
Layers within layers
February 25, 2025 at 1:16 PM
Layers within layers
Finally, we performed a sensitivity analysis using an automated feature extraction package (clinicalml.github.io/omop-learn/) vs using domain-expert knowledge for model specification.
Results were similar, which is cool bc it lowers the barriers to performing these analyses more generally
(6/7)
Results were similar, which is cool bc it lowers the barriers to performing these analyses more generally
(6/7)
February 1, 2025 at 3:28 PM
Finally, we performed a sensitivity analysis using an automated feature extraction package (clinicalml.github.io/omop-learn/) vs using domain-expert knowledge for model specification.
Results were similar, which is cool bc it lowers the barriers to performing these analyses more generally
(6/7)
Results were similar, which is cool bc it lowers the barriers to performing these analyses more generally
(6/7)
Adverse event (AE) rates were the same, if not slightly better, for first-line antibiotics vs both second-line & alternative treatments. The one exception is a slight increase in skin-related AEs (ie rash); perhaps related to sulfa allergy?
(5/7)
(5/7)
February 1, 2025 at 3:28 PM
Adverse event (AE) rates were the same, if not slightly better, for first-line antibiotics vs both second-line & alternative treatments. The one exception is a slight increase in skin-related AEs (ie rash); perhaps related to sulfa allergy?
(5/7)
(5/7)
First line treatments performed well vs second line treatments and better than alternatives (code for beta-lactams) with respect to primary outcomes (ie revisit within a month)
(4/7)
(4/7)
February 1, 2025 at 3:28 PM
First line treatments performed well vs second line treatments and better than alternatives (code for beta-lactams) with respect to primary outcomes (ie revisit within a month)
(4/7)
(4/7)
Here's our analytic method. We sought to adjust for differences that impact the likelihood of treatment and the likelihood of being lost to follow up using models optimized with ML algorithms
(3/7)
(3/7)
February 1, 2025 at 3:28 PM
Here's our analytic method. We sought to adjust for differences that impact the likelihood of treatment and the likelihood of being lost to follow up using models optimized with ML algorithms
(3/7)
(3/7)
Here's our study design. We defined time 0 as the time an individual received treatment for UTI
(2/7)
(2/7)
February 1, 2025 at 3:28 PM
Here's our study design. We defined time 0 as the time an individual received treatment for UTI
(2/7)
(2/7)
