Machine learning methods applied to audit of surgical margins after curative surgery for facial (non-melanoma) skin cancer.

We aimed to build a model to predict positive margin status after curative excision of facial non-melanoma skin cancer based on known risk factors that contribute to the complexity of the case mix. A pathology output of consecutive histology reports was requested from three oral and maxillofacial units in the south east of England. The dependent variable was a deep margin with peripheral margin clearance at a 0.5 mm threshold. A total of 3354 cases were analysed. Positivity of either the peripheral or deep margin for both squamous cell carcinoma (SCC) and basal cell carcinoma (BCC) was 15.4% at Unit 1, 21.1% at Unit 2, and 15.4% at Unit 3. Predictive models accounting for patient and tumour factors were developed using automated machine learning methods. The champion models demonstrated good discrimination for predicting margin status after excision of BCCs (AUROC = 0.67) and SCCs (AUROC = 0.71). We demonstrate that rates of positive excision margins of facial non-melanoma skin cancer (fNMSC), when adjusted by the risk prediction model, can be used to compare unit performance fairly once variations in tumour factors and patient factors are accounted for.

Exploring assessment of balance using virtual reality in patients at risk of chemotherapy-induced peripheral neuropathy.

BACKGROUND: Chemotherapy-induced peripheral neuropathy (CIPN) is a common dose-limiting toxicity for people treated for cancer. Impaired balance and falls are functional consequences of CIPN. Virtual reality (VR) technology may be able to assess balance and identify patients at risk of falls. AIMS: To assess the impact of potentially neurotoxic chemotherapy on balance using VR, and explore associations between VR balance assessment, falls and CIPN. METHODS: This prospective, repeated measures longitudinal study was conducted at two Australian cancer centres. Eligible participants were commencing adjuvant chemotherapy containing a taxane for breast cancer, or oxaliplatin for colorectal cancer (CRC), per institutional guidelines. Balance assessments using VR were conducted at baseline, end of chemotherapy and 3 and 6 months after completion of chemotherapy. Participants also completed a comprehensive CIPN assessment comprising clinical and patient-reported outcomes, and recorded falls or near falls. RESULTS: Out of 34 participants consented, 24 (71%) had breast cancer and 10 (29%) had CRC. Compared to baseline, balance threshold was reduced in 10/28 (36%) evaluable participants assessed at the end of chemotherapy, and persistent in 7/22 (32%) at 6 months. CIPN was identified in 86% at end of chemotherapy and persisted to 6 months after chemotherapy completion in 73%. Falls or near falls were reported by 12/34 (35%) participants, and were associated with impaired VR balance threshold (P = 0.002). CONCLUSIONS: While VR balance assessment was no better at identifying CIPN than existing measures, it is a potential surrogate method to assess patients at risk of falls from CIPN.

Machine learning methods applied to risk adjustment of cumulative sum chart methodology to audit free flap outcomes after head and neck surgery.

We describe a risk adjustment algorithm to benchmark and report free flap failure rates after immediate reconstruction of head and neck defects. A dataset of surgical care episodes for curative surgery for head and neck cancer and immediate reconstruction (n = 1593) was compiled from multiple NHS hospitals (n = 8). The outcome variable was complete flap failure. Classification models using preoperative patient demographic data, operation data, functional status data and tumour stage data, were built. Machine learning processes are described to model free flap failure. Overall complete flap failure was uncommon (4.7%) with a non-statistical difference seen between hospitals. The champion predictive model had acceptable discrimination (AUROC 0.66). This model was used to risk-adjust cumulative sum (CuSUM) charts. The use of CuSUM charts is a viable way to monitor in a 'Live Dashboard' this quality metric as part of the quality outcomes in oral and maxillofacial surgery audit.

Validation of a post operative complication risk prediction algorithm in a non-head and neck squamous cell carcinoma cohort.

Risk-adjusted algorithms in surgical audit attempt to adjust for patient case mix and complexity in order that published outcomes fairly reflect surgical performance and quality of care. Such risk-adjustment models have applied to head and neck squamous cell carcinoma (HNSCC). We test one algorithm, currently embedded in the oncology and reconstruction dataset within the Quality and Outcomes in Oral and Maxillofacial (QOMS) Audit, which is an artificial neural network, for its predictive accuracy on a surgical cohort receiving curative surgery for non-HNSCC pathology. A single centre retrospective case note audit of post operative complications between 2010 and 2020 was conducted on patients having curative surgery for non-HNSCC pathology. The observed complication rate was compared to the predicted probability of complications in order to test the performance of the complication risk-adjustment model. Of 1591 non-HNSCC patients, 58 met the inclusion criteria with a 30-day complication rate of 8/58 (13%). The artificial neural network predicted a complication rate of 20/58 (27%). Sensitivity (0.75), specificity (0.72) and overall accuracy (0.72) suggest acceptable discrimination. Hosmer-Lemershow Goodness of Fit test was good (p = 0.55) suggesting acceptable calibration though over-prediction of complication rate in the highest risk patents was observed. This external validation series suggests the algorithm can be applied to the non-HNSCC cohort, though some refinement of the algorithm is required to account for over-prediction of complications for higher-risk patients. With further analysis a robust means of risk adjusting for the non-HNSCC cohort should be possible.

Centralised pathology service for sentinel node biopsy in oral cavity cancer: The Southeast England Consortium experience.

BACKGROUND: Sentinel lymph node biopsy is an increasingly recognised option for accurate staging and subsequent management of the clinically negative neck in early stage oral cavity squamous cell carcinoma. However, the technique is currently underused due to several logistic constraints including increased burden on pathology services. Here, we describe the feasibility of an outsourced centralised pathology processing and reporting service for sentinel lymph node biopsies in oral cavity squamous cell carcinoma. PATIENTS AND METHODS: The Southeast England Consortium comprises four surgical centres utilising a central pathology service. Consecutive cases between January 2016 and February 2020 were retrospectively evaluated for survival outcomes and laboratory turnaround times. RESULTS: Twenty-eight per cent from a cohort of 139 patients had positive sentinel nodes. There was a trend towards greater overall, disease-free and disease-specific survival (OS, DFS and DSS, respectively) in sentinel node negative compared to sentinel node positive patients, but these differences were not statistically significant. The sensitivity, negative predictive value and false negative rate were 92.8%, 97.0% and 6.8%, respectively. The mean and mode laboratory TAT were 5 and 4 working days, respectively. CONCLUSION: An outsourced centralised pathology service is a feasible option to widen the availability of sentinel node biopsy in oral cavity squamous cell carcinoma.

Developing a risk stratification tool for audit of outcome after surgery for head and neck squamous cell carcinoma.

BACKGROUND: Patients treated surgically for head and neck squamous cell carcinoma (HNSCC) represent a heterogeneous group. Adjusting for patient case mix and complexity of surgery is essential if reporting outcomes represent surgical performance and quality of care. METHODS: A case note audit totaling 1075 patients receiving 1218 operations done for HNSCC in 4 cancer networks was completed. Logistic regression, decision tree analysis, an artificial neural network, and Naïve Bayes Classifier were used to adjust for patient case-mix using pertinent preoperative variables. RESULTS: Thirty-day complication rates varied widely (34%-51%; P < .015) between units. The predictive models allowed risk stratification. The artificial neural network demonstrated the best predictive performance (area under the curve [AUC] 0.85). CONCLUSION: Early postoperative complications are a measurable outcome that can be used to benchmark surgical performance and quality of care. Surgical outcome reporting in national clinical audits should be taking account of the patient case mix.

Is benchmarking possible in audit of early outcomes after operations for head and neck cancer?

There is a need for a validated means of adjusting for case mix in morbidity audits of patients with cancer of the head and neck. To address this, we did a multicentre audit of 3 U.K. NHS cancer networks that treat patients with head and neck cancer, to compare the incidence of early adverse postoperative outcomes and to develop a means of adjusting for case mix. We did a retrospective and prospective audit of the case notes of 901 consecutive patients who had 1034 operations for squamous cell carcinoma (SCC) of the head and neck under general anaesthesia at 3 NHS hospitals. Analysis of raw data showed postoperative 30-day mortality (n=17) to be consistent between sites (1.7%-1.9%) but 30-day complication rates varied more (34%-49%). Logistic regression models predicting morbidity discriminated well (area under the curve 0.74-0.76). Adjusted morbidity rates for the 3 units were compared on a funnel plot with 95% and 99% confidence intervals to account for random variation. It is possible to benchmark surgical performance by focusing on early postoperative outcomes in head and neck surgery. Morbidity is common and usually has a considerable impact on recovery, bed occupancy, cost, and the patient's perception of the quality of care.

Utility of a generic risk prediction score in predicting outcomes after orofacial surgery for cancer.

Understanding how variations in individual patients and perioperative events alter risk is fundamental to a useful audit of outcome. Many surgical scoring systems have been developed to allow for case-mix adjustment when reporting outcome data. To our knowledge this is the first report of the use of the physiological and operative severity score for the enumeration of mortality and morbidity (POSSUM and P-POSSUM) by a maxillofacial unit. We audited 360 operations in 245 patients between 2005 and 2008. The POSSUM morbidity equation showed reasonable discrimination (C Statistic 0.74), as did the P-POSSUM mortality equation (C Statistic 0.75), but neither showed a significant degree of goodness-of-fit (morbidity p=0.0001; mortality p=0.019). Analysis of subgroups of all elective and major operations showed no improvement in the accuracy of scores used to identify risk of complications. The need for a standard of care index for the audit of mortality and morbidity in major head and neck surgery has not yet been met.

Aortoesophageal fistula: a case report.

Patients with impacted foreign bodies in the upper aerodigestive tract present commonly to ENT clinics. This case report highlights two important issues in the management of these patients. First, if the evidence of esophageal perforation is strong and contrast swallow is negative, the physician must consider further imaging, such as contrast computed tomography. Second, ENT physicians must beware of the complications of esophageal trauma, including major vascular injury and aortoesophageal fistula, in patients with retained sharp foreign bodies in the mid-esophagus.