Cost Consequences for the NHS of Using a Two-Step Testing Method for the Detection of Clostridium difficile with a Point of Care, Polymerase Chain Reaction Test as the First Step

Clostridium difficile infection (CDI) is a common healthcare-associated infection. Current practice for diagnosing CDI in the Newcastle upon Tyne Hospitals NHS Foundation Trust involves a three-step, laboratory testing strategy using glutamate dehydrogenase (GDH) enzyme immunoassay (EIA), followed by a polymerase chain reaction (PCR) test then a toxin EIA. However, a PCR point of care test (POCT) for the C. difficile tcdB gene for screening suspected CDI cases, may provide a more efficient way of facilitating an equally effective, two-step, testing strategy with a toxin EIA. This study evaluated the cost consequences of changing from the three-step to a two-step testing strategy. A cost-consequences model was developed to compare the costs and consequences of the two strategies. Uncertainties in the model inputs were investigated with one- and two-way sensitivity analysis. The two-step, POCT strategy was estimated to save £283,282 per 1000 hospitalized NHS patients with suspected infectious diarrhea. Sensitivity analysis indicated that the turnaround time for the POCT was the largest driver for cost savings. Providing the POCT has sufficiently high diagnostic accuracy for detecting C. difficile, the two-step, POCT strategy for CDI identification is likely to be cost saving for NHS hospitals with an offsite laboratory.

Supplementary Material S1. Validation for NuTH change in C. diff algorithm: PCR prior to Toxin Testing Following a C. diff test being incorrectly tested for PCR at the same time as toxin it was noted that the toxin test was positive despite the PCR test being negative. This raised questions over the validity of results produced using the current algorithm (positive GDH followed by toxin testing and a PCR only if the toxin is proven to be negative in order to identify carriers of C. diff). The toxin positive mandatory surveillance list was then reviewed and it was identified that approximately 10% of all toxin positive specimens sent for ribotyping did not yield growth of the organism and could therefore be a false positive result. Reporting false positive cases could have adverse outcomes for both the patient who could be unnecessarily treated and also the Trust as national targets are set for these infections with financial and reputational implications if not met.
It was agreed that a PCR should be ran on any GDH positive specimens and only if positive, a toxin test should follow. This would ensure that any toxin positive results obtained are a result of production via the toxin gene detected by PCR. The additional cost of running a PCR test on all GDH positives specimens was outweighed by the financial cost of treating patients incorrectly and penalties for not meeting national targets.
It was agreed that the testing algorithm would be implemented prior to validation and that this would be carried out prospectively. This was to ensure that positive results obtained were proven to be accurate via molecular testing in a timely manner. C. diff testing was carried out using the new testing algorithm (see Error! Reference source not found., below). C. diff toxin positive specimens may be sent for ribotyping if identified on part I or part II of a patient's death certificate or if there are a cluster of cases. The first 10 C. difficile toxin positive specimens diagnosed using the new testing algorithm, requiring ribotyping would be used to validate the testing algorithm in place. Specimens were sent to the Leeds CDRN laboratory where they were cultured for C. diff and subsequently ribotyped. Results were then made available on the CDRN website and the presence of an identified ribotype was used to confirm correct identification of the C. diff case using GDH followed by PCR (Cepheid) and then toxin testing. Figure S1. Novel C. diff testing algorithm under investigation in the NuTH laboratories. 10 specimens testing positive for C. diff using PCR on the Cepheid were confirmed to be positive by ribotyping at the Leeds CDRN: Toxin B positive R005 The new testing algorithm for C. difficile implemented on 1/8/17 is fit for purpose for the diagnosis of C. diff. Performing PCR prior to toxin testing ensures that any toxin positive results obtained are due to the production of toxin from genes detected. This, in turn, will reduce the possibility of false positive toxin results ensuring that patient treatment and Trust figures are accurate.
S2. Model structure Figure S2. Simplified decision tree to assess costs and consequences of POCT strategy for CDI compared with laboratory testing strategy.  Figure S3. Care pathway of the laboratory testing strategy.  Figure S4. Proposed care pathway for the POCT strategy.

S4. Clinical interviews
Semi-structured in-depth interviews (13 in total) were conducted between January and November 2018, across the following NHS sites: Newcastle, Gateshead, Sunderland, Middlesbrough, Manchester and London. Participants were purposively selected according to their role and involvement in the diagnosis and treatment of C. diff and included infectious disease consultants, microbiologists, biomedical scientists, senior nurse managers and practitioners as well as Public Health England community consultants (see Error! Reference source not found. below, for details of distribution of roles). Two researchers (WJ and JA) conducted the fieldwork. Participants were provided with information sheets in advance, and consent forms signed prior to the start of the interviews. All interviews were digitally recorded, anonymized and transcribed in full. Interviews were typically around 60 min in length and conducted on an individual, face-to-face basis or over the phone.
Transcribed interview data were analysed using thematic analysis to generate category systems and repeated themes. Emerging themes were developed in an iterative and inductive way, breaking down and reassembling the data through a coding process. Two members of the research team (WJ and JA) undertook the analysis of the interview data. These were then reviewed and discussed at wider research team meetings, with any discrepancies resolved through this process.

S5. Literature search methods
The search strategy was designed by an experienced information specialist (FB), in collaboration with the project team. It aimed to identify existing economic models addressing the diagnosis or care pathway of patients with C. diff or other sources of diarrhoea. We conceptualised the search as: A: [C. diff OR diarrhoea] AND B: [economic models] For element A we identified thesaurus headings that described C. diff and diarrhoea. We did not use title and abstract terms because 'diarrhoea' is a term that is in common usage in many irrelevant studies so it would have made the search unmanageable, and the aim was not to conduct a comprehensive systematic review. For element B we used a search filter, a search strategy that has been tested and aims to find particular type of study. We used the Canadian Agency for Drugs and Technologies in Health (CADTH) search filter for economic evaluations and selected the section focused on economic models.
We  The average length of stay in hospital for patients with diarrhoea, CDI and other gastroenteritiscausing pathogens were obtained from NHS HRG data 2014/2015 2 as 4, 19 and 5 days respectively, and time to sample collection, before diagnostic testing, was estimated as 0.5 days by expert opinion (see Section S4). Transport time for the sample to reach the laboratory was extracted from 2014/2015 audit data from the NuTH, which encompasses two hospitals at separate sites (3 miles apart), served by a single microbiology laboratory situated at one of these sites. The common probabilities used in the model are shown in Error! Reference source not found.S5, below.  Per patient expected cost of the POCT strategy (£)