Microsoft powerpoint - escmid 03 - bugs in data 2 poster

BUGS IN THE DATA 2: Effects of identification, repeat and screening isolates on resistance estimates ML Heginbothom, JT Magee, JL Bell, AJ Howard. National Public Health Service for Wales, Abton House, Wedal Road, Cardiff, CF14 3QX, UK
Abstract: An all-Wales database, comprising details for ca. 300,000 community isolates of urinary coliforms
Conclusions: The effects of regarding heterogeneous groups of organisms as a single class were clear
Table 2. The effect of duplicate isolates and screening swab isolates on estimates of
and Escherichia coli, Staphylococcus aureus, Haemophilus influenzae, Streptococcus pneumoniae and in the analyses of urinary coliforms identified as E. coli, with those isolates reported as E. coli being antibiotic resistance
Streptococcus pyogenes was analysed. The effects of duplicates isolates, screening is olates and isolate significantly less resistant than those reported as coliforms. Species-level identification is becoming a pre- identification upon resistance estimates were investigated. Duplicates and screening isolates had negligible requisite for interpretation of susceptibility tests for disc sensitivities in the BSAC scheme1 and in many Percentage resistance ( 95% confidence interval) for;
affects on resistance estimates. However, urinary isolates reported as E. coli showed greater susceptibility than automated systems. This has advantages for surveillance, and for individual patients, where identification of a Antibiotic
All S. aureus
MSSA only
MRSA only
those that were reported as ‘coliforms’.
urinary isolate as a ‘coliform’ other than E. coli carries implications about the underlying pathology. Lack ofspecies identification, at least to the level of discriminating E. coli, may well be a false economy.
All isolates
89.5 (89.3-89.7)
87.2 (86.9-87.5)
Introduction: There is a lack of confidence in aggregate data for antibiotic resistance that discourages
Repeat and screening isolates showed little effect on community resistance estimates. This confirms results local surveillance, resulting in criticism that the data is subject to a range of potential biasing factors. These from previous studies.2-4 However, the inclusion of screening isolates of MRSA affects both methicillin- include non-standardised antimicrobial susceptibility testing (AST); inclusi on of resistance estimates based on resistance estimates for S. aureus, and community MRSA prevalence estimates. It appears unnecessary to selective testing; inclusion of estimates based on duplicate isolates and screening isolates; and failure to All isolates
20.1 (19.9-20.4)
remove duplicates isolates for community resistance surveillance, and simple rules based upon specimen site can be used to eliminate screening isolates, removing this source of bias. Extensive statistical analyses to find a break in the distribution for frequency of time intervals between duplicate isolates were unsuccessful, Purpose: To elucidate the effects of repeat isolates, screening isolates, and species identification policies for
indicating that there is no non-arbitrary time threshold beyond which matching isolates can no longer be urinary coliforms on estimates of community antibiotic resistance from routine data.
All isolates
22.1 (21.8-22.3)
8.4 (8.2-8.6)
83.9 (83.3-84.4)
regarded as duplicates. Shannon et al .4 also failed to find non-arbitrary time thresholds for duplicate systemic hospital isolates, which suggests further investigation of this aspect may not be productive.
Method: Retrospective resistance data on routine diagnostic community isolates for 1996 -2001 were
collected from fourteen Wales & ‘The Borders’ laboratories and analysed in Excel and SPSS. Duplicates 5.3 (5.1-5.5)
5.0 (4.8-5.3)
5.7 (5.3-6.1)
isolates were defined as isolates of the same species, from the same patient, with matching susceptibility patterns, but ignoring mismatches involving intermediate and untested results. Screening isolates were defined Table 3. Resistance estimates for urinary isolates reported as E. coli compared
as those overtly noted as such on the request form details, together with those from sites often included in MRSA screening but not normally involved in infection: nose, axilla and groin. Participating laboratories were with those reported as ‘coliform’; excluding non-lactose fermenting coliforms
All isolates
16.7 (16.3-17.0)
18.3 (17.9-18.7)
12.2 (11.6-12.7)
contacted to determine the method of identification (biochemical or morphological) for urinary isolates reported Percentage resistance ( 95% confidence interval) for;
Antibiotic
E. coli
E. coli
Undifferentiated
(biochemical ID)
(morphological ID)
Coliforms
Table 1. The effect of duplicate isolates on estimates of antibiotic resistance
Results: For all species investigated, exclusion of duplicate isolates (same patient, same susceptibility pattern
Percentage resistance (95% confidence interval) for;
ignoring mismatches for intermediate or not-tested results) produced negligible differences in resistance estimates Antibiotic
Coliforms
H. influenzae
St. pneumoniae
St. pyogenes
(see Table 1 and Table 2). Further restriction of this definition to matching organisms isolated within an x day-
interval also yielded negligible differences for all values of x tested (2, 4, 8, 16. 32, 64, 128, 256, 512, 1024 and 2048 days; the latter included all duplicates isolated over the 5 years survey).
Similarly, resistance estimates were closely similar when datasets for MSSA, MRSA and S. aureus that included all isolates, excluded screens, excluded duplicate isolates, or excluded both screening and duplicate isolateswere analysed. Exclusion of MRSA isolates from specimens overtly submitted as carriage screens, and from sites normally associated with carriage rather than infection, produced negligible differences in susceptibility estimates for MRSA (see Table 2). However, exclusion of repeat and screening isolates affected estimates of incidence,
and of methicillin resistance in S. aureus (20.1% resistance to flucloxacillin (methicillin) for all S. aureus isolates compared to 15.5% resistance excluding duplicates and screens).
Laboratory identification of urinary coliform isolates varied between the fourteen participating laboratories.
References:
Isolates identified biochemically (minimum of indole tests and lactose fermentation) as E. coli were significantly 1. Andrews, J. for the BSAC Working Party Report on Susceptibility Testing. (2001). BSAC standardized disc less resistant than those identified by colony morphology (morphological identification did not include the use of susceptibility testing method. Journal of Antimicrobial Chemotherapy 48, Suppl. 1, 43-57.
chromogenic agar), and both were significantly less resistant than urinary coliforms with no species identification 2. Huovinen, P. (1985). Recording of antimicrobial resistances of urinary tract isolates - effect of repeat (see Table 3). For example, resistance to co-amoxiclav for biochemically identified E. coli was 5.1% compared to
samples on resistance levels. Journal of Antimicrobial Chemotherapy 16, 443-447.
8.3% resistance for those identified morphologically and 9.7% resistance for undifferentiated coliforms.
3. Howard, A.J., Magee, J.T. et al. (2001) Factors associated with antibiotic resistance in colifo rm organisms from community urinary tract infections in Wales. Journal of Antimicrobial Chemotherapy 47, 305-313.
Please note that data for non-lactose fermenting coliforms were excluded from this study.
4. Shannon, K.P. & French, G.L. (2002). Antibiotic resistance: effect of different criteria for classifying isolates as duplicates on apparent resistance frequencies. Journal of Antimicrobial Chemotherapy 49, 201-204.

Source: http://www.wales.nhs.uk/sites3/Documents/457/ESCMID%2003%20-%20Bugs%20in%20Data%202%20Poster.pdf