Monash EBS is hiring

Carbapenemase-producing Enterobacterales

  • Gram-negative bacilli occurring naturally in the GI tract
  • Resistant to carbapenem antibiotics
  • Several different carbapenemases genes in CPE:
    • Imipenemase (IMP)
    • Klebsiella penumoniae carbapenemase (KPC)
    • New-Delhi metallo-β-lactamase (NDM)
    • Verona integron-encoded metallo-β-lactamase (VIM) and
    • Oxacillinases (OXA)
  • For Enterobacterales, IMP(-4) most commonly found in Victoria (28% of all cases)
  • Assumed to be colonised indefinitely

Carbapenemase-producing Enterobacterales

  • Surveillance officially established in Victoria in 2015, became notifiable in 2018
  • Since becoming notifiable, notifications have been steadily increasing

Modelling possible control

  • We know that healthcare is a network phenomenon
  • Need to understand if CPE+ patients are different to CPE- patients

Health Services in Victoria, Australia

  • Health services are individually governed
  • Centralised guidelines for reporitng
  • Have their own IP results, testing/screening requirements, admission & transfer patterns
  • Are often spread across multiple campuses which operate independently but are administered centrally

As these services tend to operate on the same patient data platform, we assume that previously diagnosed patients will have infection prevention protocols in place if they are re-admitted to the same service.

This is less likely to be true in a different health service, as there is no centralised notification system to other health services.

Notification system

Data sources

Victorian Admitted Episodes Dataset (VAED):

  • Line list of every hospital admission (public and private)
  • Reports demographics, conditions, treatments, timing, location

Public Health Event Surveillance System (PHESS):

  • Records information on notifiable disease events
  • Includes patient demographics, date of notification, date of test
  • Operational system, used for contact tracing and outbreak analytics

Network construction

We model how patients move around the network.

  • Nodes: Hospital (physical)
  • Edge: Transfer of patient between two hospitals, within 365 days
  • Weight: Number of patients transferred

This is a weighted, directed network. The worst kind of non-temporal network.

Network construction

  • Clustering: Infomap (based on a random walk through the network)

\max_{M} L(M) = q H(\mathcal{Q}) + \sum_{i=1}^m p^{i} H(\mathcal{P}^i)

Effectively maximising entropy of the “walk” subject to a partition, Q, weighting within and between cluster movements.

p and q informed by edge weights

Network construction

  • Importance of certain facilites: Adjusted Rand Index

R = \frac{a+b}{a+b+c+d} = \frac{a+b}{n\choose 2},

where

  • a is the number of pairs of elements that are in the same subset in X and the same subset in Y,
  • b be the number of pairs of elements that are in different subsets in X and different subsets of Y,
  • c be the number of pairs of elements that are in the same subset in X and different subsets in Y and,
  • d be the number of pairs of elements that are in different subsets in X and the same subset in Y.

Network exploration

As close to results as we’re gonna get

Admission demographics

Variable N
Gender
Male 12,723,729
Female 14,071,488
Other 1,190
Age Group
<20 2,567,643
20-39 4,744,744
50-59 6,372,288
>60 13,111,732
Length of Stay
Mean 2.739
Median (IQR) 2 (1-3)

Network exploration

  • Density: High (0.44)
  • Clustering: Reasonably strong (foreshadowing)

Network exploration

Clustering

Clustering

Clustering

Node importance

ARI: Interpret at proportion of clustering structure “preserved”

  • Removing small and same-day facilities gives an ARI of between 0.58 and 0.62
    • Seems key to making connections between larger hospitals
    • Removing one-by-one or all doesn’t notably change ARI
  • Removing public hospitals gives ARI of 0.22
  • Removing private hospitals gives ARI of 0.58

Next steps

  • Hospital system is dense, geographically clustered

  • Transfers are incredibly frequent, tend to move towards metropolitan Melbourne

  • Need to include pretty much all facilities in system consideration

  • Surveillance should be considered geographically

    • Local public health units?

  • Impact of patients going home

  • Testing/sampling for conditions

This is a very extensive and detailed dataset, with many more project ideas than time.

But presents an exciting opportunity to build a data-driven surveillance and outbreak response system.