Explaining Predictive Model Performance: An Experimental Study of Data Preparation and Model Choice.

Although confirmatory modeling has dominated much of applied research in medical, business, and behavioral sciences, modeling large data sets with the goal of accurate prediction has become more widely accepted. The current practice for fitting predictive models is guided by heuristic-based modeling frameworks that lead researchers to make a series of often isolated decisions regarding data preparation and cleaning that may result in substandard predictive performance. In this article, we use an experimental design to evaluate the impact of six factors related to data preparation and model selection (techniques for numerical imputation, categorical imputation, encoding, subsampling for unbalanced data, feature selection, and machine learning algorithm) and their interactions on the predictive accuracy of models applied to a large, publicly available heart transplantation database. Our factorial experiment includes 10,800 models evaluated on 5 independent test partitions of the data. Results confirm that some decisions made early in the modeling process interact with later decisions to affect predictive performance; therefore, the current practice of making these decisions independently can negatively affect predictive outcomes. A key result of this case study is to highlight the need for improved rigor in applied predictive research. By using the scientific method to inform predictive modeling, we can work toward a framework for applied predictive modeling and a standard for reproducibility in predictive research.

Explaining the Varying Patterns of COVID-19 Deaths Across the United States: 2-Stage Time Series Clustering Framework.

BACKGROUND: Socially vulnerable communities are at increased risk for adverse health outcomes during a pandemic. Although this association has been established for H1N1, Middle East respiratory syndrome (MERS), and COVID-19 outbreaks, understanding the factors influencing the outbreak pattern for different communities remains limited. OBJECTIVE: Our 3 objectives are to determine how many distinct clusters of time series there are for COVID-19 deaths in 3108 contiguous counties in the United States, how the clusters are geographically distributed, and what factors influence the probability of cluster membership. METHODS: We proposed a 2-stage data analytic framework that can account for different levels of temporal aggregation for the pandemic outcomes and community-level predictors. Specifically, we used time-series clustering to identify clusters with similar outcome patterns for the 3108 contiguous US counties. Multinomial logistic regression was used to explain the relationship between community-level predictors and cluster assignment. We analyzed county-level confirmed COVID-19 deaths from Sunday, March 1, 2020, to Saturday, February 27, 2021. RESULTS: Four distinct patterns of deaths were observed across the contiguous US counties. The multinomial regression model correctly classified 1904 (61.25%) of the counties' outbreak patterns/clusters. CONCLUSIONS: Our results provide evidence that county-level patterns of COVID-19 deaths are different and can be explained in part by social and political predictors.

Modeling the differences in the time-series profiles of new COVID-19 daily confirmed cases in 3,108 contiguous U.S. counties: A retrospective analysis.

OBJECTIVE: The COVID-19 pandemic in the U.S. has exhibited a distinct multiwave pattern beginning in March 2020. Paradoxically, most counties do not exhibit this same multiwave pattern. We aim to answer three research questions: (1) How many distinct clusters of counties exhibit similar COVID-19 patterns in the time-series of daily confirmed cases? (2) What is the geographic distribution of the counties within each cluster? and (3) Are county-level demographic, socioeconomic and political variables associated with the COVID-19 case patterns? MATERIALS AND METHODS: We analyzed data from counties in the U.S. from March 1, 2020 to January 2, 2021. Time series clustering identified clusters in the daily confirmed cases of COVID-19. An explanatory model was used to identify demographic, socioeconomic and political variables associated with the outbreak patterns. RESULTS: Three patterns were identified from the cluster solution including counties in which cases are still increasing, those that peaked in the late fall, and those with low case counts to date. Several county-level demographic, socioeconomic, and political variables showed significant associations with the identified clusters. DISCUSSION: The pattern of the outbreak is related both to the geographic location within the U.S. and several variables including population density and government response. CONCLUSION: The reported pattern of cases in the U.S. is observed through aggregation of the daily confirmed COVID-19 cases, suggesting that local trends may be more informative. The pattern of the outbreak varies by county, and is associated with important demographic, socioeconomic, political and geographic factors.

Perceptions of UK clinicians towards postoperative critical care.

Postoperative critical care is a finite resource that is recommended for high-risk patients. Despite national recommendations specifying that such patients should receive postoperative critical care, there is evidence that these recommendations are not universally followed. We performed a national survey aiming to better understand how patients are risk-stratified in practice; elucidate clinicians' opinions about how patients should be selected for critical care; and determine factors which affect the actual provision of postoperative critical care. As part of the second Sprint National Anaesthesia Project, epidemiology of critical care after surgery study, we distributed a paper survey to anaesthetists, surgeons and intensivists providing peri-operative care during a single week in March 2017. We collected data on respondent characteristics, and their opinions of postoperative critical care provision, potential benefits and real-world challenges. We undertook both quantitative and qualitative analyses to interpret the responses. We received 10,383 survey responses from 237 hospitals across the UK. Consultants used a lower threshold for critical care admission than other career grades, indicating potentially more risk-averse behaviour. The majority of respondents reported that critical care provision was inadequate, and cited the value of critical care as being predominantly due to higher nurse: patient ratios. Use of objective risk assessment tools was poor, and patients were commonly selected for critical care based on procedure-specific pathways rather than individualised risk assessment. Challenges were highlighted in the delivery of peri-operative critical care services, such as an overall lack of capacity, competition for beds with non-surgical cases and poor flow through the hospital leading to bed 'blockages'. Critical care is perceived to provide benefit to high-risk surgical patients, but there is variation in practice about the definition and determination of risk, how patients are referred and how to deal with the lack of critical care resources. Future work should focus on evaluating 'enhanced care' units for postoperative patients, how to better implement individualised risk assessment in practice, and how to improve patient flow through hospitals.

A forecasting framework for predicting perceived fatigue: Using time series methods to forecast ratings of perceived exertion with features from wearable sensors.

Advancements in sensing and network technologies have increased the amount of data being collected to monitor the worker conditions. In this study, we consider the use of time series methods to forecast physical fatigue using subjective ratings of perceived exertion (RPE) and gait data from wearable sensors captured during a simulated in-lab manual material handling task (Lab Study 1) and a fatiguing squatting with intermittent walking cycle (Lab Study 2). To determine whether time series models can accurately forecast individual response and for how many time periods ahead, five models were compared: naïve method, autoregression (AR), autoregressive integrated moving average (ARIMA), vector autoregression (VAR), and the vector error correction model (VECM). For forecasts of three or more time periods ahead, the VECM model that incorporates historical RPE and wearable sensor data outperformed the other models with median mean absolute error (MAE) <1.24 and median MAE <1.22 across all participants for Lab Study 1 and Lab Study 2, respectively. These results suggest that wearable sensor data can support forecasting a worker's condition and the forecasts obtained are as good as current state-of-the-art models using multiple sensors for current time prediction.

Regional, socioeconomic and behavioural- impacts on consumer acceptability of beef in Northern Ireland, Republic of Ireland and Great Britain.

This study was commissioned to assess if there are regional differences in the acceptability of beef between consumers from Northern Ireland (NI), Republic of Ireland (ROI) and Great Britain (GB). Palatability traits were affected by socioeconomic and behavioural factors such as preferred cooking endpoint, animal welfare, value, health aspects of beef product, ease of preparation as well as consumption frequency for specific cuts. "Willingness to pay" (WTP) was influenced by income, preferred cooking endpoint, value of beef product, ease of preparation and consumption frequency for frying steak. Results showed that GB consumers scored higher for the same striploin steak compared to NI and ROI consumers. This may be due to differences in the motivation for beef choice and/or consumption habits. GB consumers were less concerned about the healthiness of beef product and beef origin. In addition, a higher consumption frequency for rump was reported in GB, which may explain the higher sensory scores observed among GB consumers for striploins.

Review: Beef-eating quality: a European journey.

This paper reviews recent research into predicting the eating qualities of beef. A range of instrumental and grading approaches have been discussed, highlighting implications for the European beef industry. Studies incorporating a number of instrumental and spectroscopic techniques illustrate the potential for online systems to non-destructively measure muscle pH, colour, fat and moisture content of beef with R 2 (coefficient of determination) values >0.90. Direct predictions of eating quality (tenderness, flavour, juiciness) and fatty acid content using these methods are also discussed though success is greatly variable. R 2 values for instrumental measures of tenderness have been quoted as high as 0.85 though R 2 values for sensory tenderness values can be as low as 0.01. Discriminant analysis models can improve prediction of variables such as pH and shear force, correctly classifying beef samples into categorical groups with >90% accuracy. Prediction of beef flavour continues to challenge researchers and the industry alike, with R 2 values rarely quoted above 0.50, regardless of instrumental or statistical analysis used. Beef grading systems such as EUROP and United States Department of Agriculture systems provide carcase classification and some indication of yield. Other systems attempt to classify the whole carcase according to expected eating quality. These are being supplemented by schemes such as Meat Standards Australia (MSA), based on consumer satisfaction for individual cuts. In Australia, MSA has grown steadily since its inception generating a 10% premium for the beef industry in 2015-16 of $187 million. There is evidence that European consumers would respond to an eating quality guarantee provided it is simple and independently controlled. A European beef quality assurance system might encompass environmental and nutritional measures as well as eating quality and would need to be profitable, simple, effective and sufficiently flexible to allow companies to develop their own brands.

Anaesthesia, surgery, and life-threatening allergic reactions: protocol and methods of the 6th National Audit Project (NAP6) of the Royal College of Anaesthetists.

BACKGROUND: Anaphylaxis during anaesthesia is a serious complication for patients and anaesthetists. METHODS: The Sixth National Audit Project (NAP6) of the Royal College of Anaesthetists examined the incidence, predisposing factors, management, and impact of life-threatening perioperative anaphylaxis in the UK. NAP6 included: a national survey of anaesthetists' experiences and perceptions; a national survey of allergy clinics; a registry collecting detailed reports of all Grade 3-5 perioperative anaphylaxis cases for 1 yr; and a national survey of anaesthetic workload and perioperative allergen exposure. NHS and independent sector (IS) hospitals were approached to participate. Cases were reviewed by a multi-disciplinary expert panel (anaesthetists, intensivists, allergists, immunologists, patient representatives, and stakeholders) using a structured process designed to minimise bias. Clinical management and investigation were compared with published guidelines. This paper describes detailed study methods and reports on project engagement by NHS and IS hospitals. The methodology includes a new classification of perioperative anaphylaxis and a new structured method for classifying suspected anaphylactic events including the degree of certainty with which a causal trigger agent can be attributed. RESULTS: NHS engagement was complete (100% of hospitals). Independent sector engagement was limited (13% of approached hospitals). We received >500 reports of Grade 3-5 perioperative anaphylaxis, with 266 suitable for analysis. We identified 199 definite or probable culprit agents in 192 cases. CONCLUSIONS: The methods of NAP6 were robust in identifying causative agents of anaphylaxis, and support the accompanying analytical papers.

Anaesthesia, surgery, and life-threatening allergic reactions: epidemiology and clinical features of perioperative anaphylaxis in the 6th National Audit Project (NAP6).

BACKGROUND: Anaphylaxis during anaesthesia is a serious complication for patients and anaesthetists. METHODS: The 6th National Audit Project (NAP6) on perioperative anaphylaxis collected and reviewed 266 reports of Grades 3-5 anaphylaxis over 1 yr from all NHS hospitals in the UK. RESULTS: The estimated incidence was ≈1:10 000 anaesthetics. Case exclusion because of reporting delays or incomplete data means true incidence might be ≈70% higher. The distribution of 199 identified culprit agents included antibiotics (94), neuromuscular blocking agents (65), chlorhexidine (18), and Patent Blue dye (9). Teicoplanin comprised 12% of antibiotic exposures, but caused 38% of antibiotic-induced anaphylaxis. Eighteen patients reacted to an antibiotic test dose. Succinylcholine-induced anaphylaxis, mainly presenting with bronchospasm, was two-fold more likely than other neuromuscular blocking agents. Atracurium-induced anaphylaxis mainly presented with hypotension. Non-depolarising neuromuscular blocking agents had similar incidences to each other. There were no reports of local anaesthetic or latex-induced anaphylaxis. The commonest presenting features were hypotension (46%), bronchospasm (18%), tachycardia (9.8%), oxygen desaturation (4.7%), bradycardia (3%), and reduced/absent capnography trace (2.3%). All patients were hypotensive during the episode. Onset was rapid for neuromuscular blocking agents and antibiotics, but delayed with chlorhexidine and Patent Blue dye. There were 10 deaths and 40 cardiac arrests. Pulseless electrical activity was the usual type of cardiac arrest, often with bradycardia. Poor outcomes were associated with increased ASA, obesity, beta blocker, and angiotensin-converting enzyme inhibitor medication. Seventy per cent of cases were reported to the hospital incident reporting system, and only 24% to Medicines and Healthcare products Regulatory Agency via the Yellow Card Scheme. CONCLUSIONS: The overall incidence of perioperative anaphylaxis was estimated to be 1 in 10 000 anaesthetics.

Cross-sectional study of perioperative drug and allergen exposure in UK practice in 2016: the 6th National Audit Project (NAP6) Allergen Survey.

BACKGROUND: Details of the current UK drug and allergen exposure were needed for interpretation of reports of perioperative anaphylaxis to the 6th National Audit Project (NAP6). METHODS: We performed a cross-sectional survey of 356 NHS hospitals determining anaesthetic drug usage in October 2016. All cases cared for by an anaesthetist were included. RESULTS: Responses were received from 342 (96%) hospitals. Within-hospital return rates were 96%. We collected 15 942 forms, equating to an annual caseload of 3.1 million, including 2.4 million general anaesthetics. Propofol was used in 74% of all cases and 90% of general anaesthetics. Maintenance included a volatile agent in 95% and propofol in 8.7%. Neuromuscular blocking agents were used in 47% of general anaesthetics. Analgesics were used in 88% of cases: opioids, 82%; paracetamol, 56%; and non-steroidal anti-inflammatory drugs, 28%. Antibiotics were administered in 57% of cases, including 2.5 million annual perioperative administrations; gentamicin, co-amoxiclav, and cefuroxime were most commonly used. Local anaesthetics were used in 74% cases and 70% of general anaesthetics. Anti-emetics were used in 73% of cases: during general anaesthesia, ondansetron in 78% and dexamethasone in 60%. Blood products were used in ≈3% of cases, gelatin <2%, starch very rarely, and tranexamic acid in ≈6%. Chlorhexidine and povidone-iodine exposures were 74% and 40% of cases, and 21% reported a latex-free environment. Exposures to bone cement, blue dyes, and radiographic contrast dye were each reported in 2-3% of cases. CONCLUSIONS: This survey provides insights into allergen exposures in perioperative care, which is important as denominator data for the NAP6 registry.

Anaesthesia, surgery, and life-threatening allergic reactions: management and outcomes in the 6th National Audit Project (NAP6).

BACKGROUND: Anaphylaxis during anaesthesia is a serious complication for patients and anaesthetists. There is little published information on management and outcomes of perioperative anaphylaxis in the UK. METHODS: The 6th National Audit Project of the Royal College of Anaesthetists (NAP6) collected and reviewed 266 reports of Grade 3-5 anaphylaxis from all UK NHS hospitals over 1 yr. Quality of management was assessed against published guidelines. RESULTS: Appropriately senior anaesthetists resuscitated all patients. Immediate management was 'good' in 46% and 'poor' in 15%. Recognition and treatment of anaphylaxis were prompt in 97% and 83% of cases, respectively. Epinephrine was administered i.v. in 76%, i.m. in 14%, both in 6%, and not at all in 11% of cases. A catecholamine infusion was administered in half of cases. Cardiac arrests (40 cases; 15%) were promptly treated but cardiac compressions were omitted in half of patients with unrecordable BP. The surgical procedure was abandoned in most cases, including 10% where surgery was urgent. Of 54% admitted to critical care, 70% were level 3, with most requiring catecholamine infusions. Ten (3.8%) patents (mostly elderly with cardiovascular disease) died from anaphylaxis. Corticosteroids and antihistamines were generally administered early. We found no clear evidence of harm or benefit from chlorphenamine. Two patients received vasopressin and one glucagon. Fluid administration was inadequate in 19% of cases. Treatment included sugammadex in 19 cases, including one when rocuronium had not been administered. Adverse sequelae (psychological, cognitive, or physical) were reported in one-third of cases. CONCLUSIONS: Management of perioperative anaphylaxis could be improved, especially with respect to administration of epinephrine, cardiac compressions, and i.v. fluid. Sequelae were common.

An observational national study of anaesthetic workload and seniority across the working week and weekend in the UK in 2016: the 6th National Audit Project (NAP6) Activity Survey.

BACKGROUND: UK national anaesthetic activity was studied in 2013 but weekend working was not examined. Understanding changes since 2013 in workload and manpower distribution, including weekends, would be of value in workforce planning. METHODS: We performed an observational survey of NHS hospitals' anaesthetic practice in October 2016 as part of the 6th National Audit Project of the Royal College of Anaesthetists (NAP6). All cases cared for by an anaesthetist during the study period were included. Patient characteristics and details of anaesthetic conduct were collected by local anaesthetists. RESULTS: Responses were received from 342/356 (96%) hospitals. In total, 15 942 cases were reported, equating to an annual anaesthetic workload of ≈3.13 million cases. Approximately 95% (9888/10 452) of elective and 72% (3184/4392) of emergency work was performed on weekdays and 89% (14 145/15 942) of activity was led by senior (consultant or career grade) anaesthetists and 1.1% (180/15942) by those with <2 yr anaesthetic experience. During weekends case urgency increased, the proportion of healthy patients reduced and case mix changed. Cases led by senior anaesthetists fell to 80% (947/1177) on Saturday and 66% (342/791) on Sunday. Senior involvement in obstetric anaesthetic activity was 69% (628/911) during the week and 45% (182/402) at weekends, compared with 93% (791/847) in emergency orthopaedic procedures during the week and 89% (285/321) at weekends. Since 2013, the proportion of obese patients, elective weekend working, and depth of anaesthesia monitoring has increased [12% (1464/12 213) vs 2.8%], but neuromuscular monitoring has not [37% (2032/5532) vs 38% of paralysed cases]. CONCLUSIONS: Senior clinicians deliver most UK anaesthesia care, including at weekends. Our findings are important for any planned workforce reorganisation to rationalise 7-day working.

Specialist perioperative allergy clinic services in the UK 2018: Results from the Royal College of Anaesthetists Sixth National Audit Project (NAP6) investigation of perioperative anaphylaxis.

BACKGROUND: The Royal College of Anaesthetists 6th National Audit Project examined Grade 3-5 perioperative anaphylaxis for 1 year in the UK. OBJECTIVE: To describe the causes and investigation of anaphylaxis in the NAP6 cohort, in relation to published guidance and previous baseline survey results. METHODS: We used a secure registry to gather details of Grade 3-5 perioperative anaphylaxis. Anonymous reports were aggregated for analysis and reviewed in detail. Panel consensus diagnosis, reaction grade, review of investigations and clinic assessment are reported and compared to the prior NAP6 baseline clinic survey. RESULTS: A total of 266 cases met inclusion criteria between November 2015 and 2016, detailing reactions and investigations. One hundred and ninety-two of 266 (72%) had anaphylaxis with a trigger identified, of which 140/192 (75%) met NAP6 criteria for IgE-mediated allergic anaphylaxis, 13% lacking evidence of positive IgE tests were labelled "non-allergic anaphylaxis". 3% were non-IgE-mediated anaphylaxis. Adherence to guidance was similar to the baseline survey for waiting time for clinic assessment. However, lack of testing for chlorhexidine and latex, non-harmonized testing practices and poor coverage of all possible culprits was confirmed. Challenge testing may be underused and many have unacceptably delayed assessments, even in urgent cases. Communication or information provision for patients was insufficient, especially for avoidance advice and communication of test results. Insufficient detail regarding skin test methods was available to draw conclusions regarding techniques. CONCLUSION AND CLINICAL RELEVANCE: Current clinical assessment in the UK is effective but harmonization of approach to testing, access to services and MHRA reporting is needed. Expert anaesthetist involvement should increase to optimize diagnostic yield and advice for future anaesthesia. Dynamic tryptase evaluation improves detection of tryptase release where peak tryptase is <14 µg/L and should be adopted. Standardized clinic reports containing appropriate details of tests, conclusions, avoidance, cross-reactivity and suitable alternatives are required to ensure effective, safe future management options.

Review: The variability of the eating quality of beef can be reduced by predicting consumer satisfaction.

The Meat Standards Australia (MSA) grading scheme has the ability to predict beef eating quality for each 'cut×cooking method combination' from animal and carcass traits such as sex, age, breed, marbling, hot carcass weight and fatness, ageing time, etc. Following MSA testing protocols, a total of 22 different muscles, cooked by four different cooking methods and to three different degrees of doneness, were tasted by over 19 000 consumers from Northern Ireland, Poland, Ireland, France and Australia. Consumers scored the sensory characteristics (tenderness, flavor liking, juiciness and overall liking) and then allocated samples to one of four quality grades: unsatisfactory, good-every-day, better-than-every-day and premium. We observed that 26% of the beef was unsatisfactory. As previously reported, 68% of samples were allocated to the correct quality grades using the MSA grading scheme. Furthermore, only 7% of the beef unsatisfactory to consumers was misclassified as acceptable. Overall, we concluded that an MSA-like grading scheme could be used to predict beef eating quality and hence underpin commercial brands or labels in a number of European countries, and possibly the whole of Europe. In addition, such an eating quality guarantee system may allow the implementation of an MSA genetic index to improve eating quality through genetics as well as through management. Finally, such an eating quality guarantee system is likely to generate economic benefits to be shared along the beef supply chain from farmers to retailors, as consumers are willing to pay more for a better quality product.

Untrained consumer assessment of the eating quality of European beef: 2. Demographic factors have only minor effects on consumer scores and willingness to pay.

The beef industry must become more responsive to the changing market place and consumer demands. An essential part of this is quantifying a consumer's perception of the eating quality of beef and their willingness to pay for that quality, across a broad range of demographics. Over 19 000 consumers from Northern Ireland, Poland, Ireland and France each tasted seven beef samples and scored them for tenderness, juiciness, flavour liking and overall liking. These scores were weighted and combined to create a fifth score, termed the Meat Quality 4 score (MQ4) (0.3×tenderness, 0.1×juiciness, 0.3×flavour liking and 0.3×overall liking). They also allocated the beef samples into one of four quality grades that best described the sample; unsatisfactory, good-every-day, better-than-every-day or premium. After the completion of the tasting panel, consumers were then asked to detail, in their own currency, their willingness to pay for these four categories which was subsequently converted to a proportion relative to the good-every-day category (P-WTP). Consumers also answered a short demographic questionnaire. The four sensory scores, the MQ4 score and the P-WTP were analysed separately, as dependant variables in linear mixed effects models. The answers from the demographic questionnaire were included in the model as fixed effects. Overall, there were only small differences in consumer scores and P-WTP between demographic groups. Consumers who preferred their beef cooked medium or well-done scored beef higher, except in Poland, where the opposite trend was found. This may be because Polish consumers were more likely to prefer their beef cooked well-done, but samples were cooked medium for this group. There was a small positive relationship with the importance of beef in the diet, increasing sensory scores by about 4% in Poland and Northern Ireland. Men also scored beef about 2% higher than women for most sensory scores in most countries. In most countries, consumers were willing to pay between 150 and 200% more for premium beef, and there was a 50% penalty in value for unsatisfactory beef. After quality grade, by far the greatest influence on P-WTP was country of origin. Consumer age also had a small negative relationship with P-WTP. The results indicate that a single quality score could reliably describe the eating quality experienced by all consumers. In addition, if reliable quality information is delivered to consumers they will pay more for better quality beef, which would add value to the beef industry and encourage improvements in quality.

Untrained consumer assessment of the eating quality of beef: 1. A single composite score can predict beef quality grades.

Quantifying consumer responses to beef across a broad range of demographics, nationalities and cooking methods is vitally important for any system evaluating beef eating quality. On the basis of previous work, it was expected that consumer scores would be highly accurate in determining quality grades for beef, thereby providing evidence that such a technique could be used to form the basis of and eating quality grading system for beef. Following the Australian MSA (Meat Standards Australia) testing protocols, over 19 000 consumers from Northern Ireland, Poland, Ireland, France and Australia tasted cooked beef samples, then allocated them to a quality grade; unsatisfactory, good-every-day, better-than-every-day and premium. The consumers also scored beef samples for tenderness, juiciness, flavour-liking and overall-liking. The beef was sourced from all countries involved in the study and cooked by four different cooking methods and to three different degrees of doneness, with each experimental group in the study consisting of a single cooking doneness within a cooking method for each country. For each experimental group, and for the data set as a whole, a linear discriminant function was calculated, using the four sensory scores which were used to predict the quality grade. This process was repeated using two conglomerate scores which are derived from weighting and combining the consumer sensory scores for tenderness, juiciness, flavour-liking and overall-liking, the original meat quality 4 score (oMQ4) (0.4, 0.1, 0.2, 0.3) and current meat quality 4 score (cMQ4) (0.3, 0.1, 0.3, 0.3). From the results of these analyses, the optimal weightings of the sensory scores to generate an 'ideal meat quality 4 score (MQ4)' for each country were calculated, and the MQ4 values that reflected the boundaries between the four quality grades were determined. The oMQ4 weightings were far more accurate in categorising European meat samples than the cMQ4 weightings, highlighting that tenderness is more important than flavour to the consumer when determining quality. The accuracy of the discriminant analysis to predict the consumer scored quality grades was similar across all consumer groups, 68%, and similar to previously reported values. These results demonstrate that this technique, as used in the MSA system, could be used to predict consumer assessment of beef eating quality and therefore to underpin a commercial eating quality guarantee for all European consumers.

European conformation and fat scores have no relationship with eating quality.

European conformation and fat grades are a major factor determining carcass value throughout Europe. The relationships between these scores and sensory scores were investigated. A total of 3786 French, Polish and Irish consumers evaluated steaks, grilled to a medium doneness, according to protocols of the ���Meat Standards Australia��� system, from 18 muscles representing 455 local, commercial cattle from commercial abattoirs. A mixed linear effects model was used for the analysis. There was a negative relationship between juiciness and European conformation score. For the other sensory scores, a maximum of three muscles out of a possible 18 demonstrated negative effects of conformation score on sensory scores. There was a positive effect of European fat score on three individual muscles. However, this was accounted for by marbling score. Thus, while the European carcass classification system may indicate yield, it has no consistent relationship with sensory scores at a carcass level that is suitable for use in a commercial system. The industry should consider using an additional system related to eating quality to aid in the determination of the monetary value of carcasses, rewarding eating quality in addition to yield.

The variation in the eating quality of beef from different sexes and breed classes cannot be completely explained by carcass measurements.

Delivering beef of consistent quality to the consumer is vital for consumer satisfaction and will help to ensure demand and therefore profitability within the beef industry. In Australia, this is being tackled with Meat Standards Australia (MSA), which uses carcass traits and processing factors to deliver an individual eating quality guarantee to the consumer for 135 different 'cut by cooking methods' from each carcass. The carcass traits used in the MSA model, such as ossification score, carcass weight and marbling explain the majority of the differences between breeds and sexes. Therefore, it was expected that the model would predict with eating quality of bulls and dairy breeds with good accuracy. In total, 8128 muscle samples from 482 carcasses from France, Poland, Ireland and Northern Ireland were MSA graded at slaughter then evaluated for tenderness, juiciness, flavour liking and overall liking by untrained consumers, according to MSA protocols. The scores were weighted (0.3, 0.1, 0.3, 0.3) and combined to form a global eating quality (meat quality (MQ4)) score. The carcasses were grouped into one of the three breed categories: beef breeds, dairy breeds and crosses. The difference between the actual and the MSA-predicted MQ4 scores were analysed using a linear mixed effects model including fixed effects for carcass hang method, cook type, muscle type, sex, country, breed category and postmortem ageing period, and random terms for animal identification, consumer country and kill group. Bulls had lower MQ4 scores than steers and females and were predicted less accurately by the MSA model. Beef breeds had lower eating quality scores than dairy breeds and crosses for five out of the 16 muscles tested. Beef breeds were also over predicted in comparison with the cross and dairy breeds for six out of the 16 muscles tested. Therefore, even after accounting for differences in carcass traits, bulls still differ in eating quality when compared with females and steers. Breed also influenced eating quality beyond differences in carcass traits. However, in this case, it was only for certain muscles. This should be taken into account when estimating the eating quality of meat. In addition, the coefficients used by the Australian MSA model for some muscles, marbling score and ultimate pH do not exactly reflect the influence of these factors on eating quality in this data set, and if this system was to be applied to Europe then the coefficients for these muscles and covariates would need further investigation.