Overall, 123 students from the five cohorts were identified as strugglers and there were 492 controls. Fifty three strugglers were initially identified as APC attendees, nine as course terminations, 30 as voluntary withdrawals, and 31 as course suspensions. Thus of the 961 students admitted over the five years, 123 (12.8%) experienced problems, including 41 (4%) who left the course before graduation, as described below.
Extent of strugglers' problems
The strugglers experienced considerable academic problems. Seventy of the 123 (57%) failed three or more preclinical exams and only 92 (75%) had graduated as BMedSci (bachelor of medical science) at the end of the third year; of the 31 remaining, 21 left the course voluntarily, eight had their course terminated, and two were still on the preclinical course. At a date when all students should have completed their training, a further 12 had left during the clinical course, eight were still on the clinical course, and only 72 (59%) of the original group of strugglers had graduated BMBS (bachelor of medicine and bachelor of surgery).
In addition, 56 of the 123 (46%) had disruptive personal or medical problems noted in their undergraduate records, including a high incidence of depressive illness (strugglers 29/123 (24%) v controls 6/492 (1%); χ2 91.6, P < 0.001). Overall, 34 (28%) of the strugglers had attitude problems noted in their undergraduate records.
Preadmission factors associated with strugglers (univariate analyses).
Tables and summarise these results. The non-academic factors that were significantly more common in the strugglers were the presence of negative comments in the head teacher's reference, the late offer of a place, male sex, and slightly older age. Non-white ethnicity was significant in UK students. There was no significant difference between UK strugglers and controls in social deprivation (Townsend score6
Table 2 Univariate (2) analyses of discontinuous non-academic and academic variables. Figures are numbers (percentages) of students
Univariate (Mann-Whitney U) analyses of continuous non-academic and academic variables
The significant academic associations with the strugglers were lower mean examination grade at GCSE, lower mean examination grade at A level, and not achieving a grade A at A level biology. Such students were also more likely to have significantly lower UCAS points (point score for top three A levels) and total tariff scores (point score for all A levels taken). There were no significant differences between the groups in terms of A level chemistry grade or having undertaken a previous degree.
Preadmission factors significantly and independently predicting strugglers (multivariate analysis)
We entered preadmission factors that had shown significant univariate associations with struggler status (above) into a binary logistic regression. shows the data for all students excluding ethnicity as an explanatory variable because it was not known for overseas students, and data for UK students including ethnicity as that was recorded for most of these.
Significant independent predictors of strugglers (binary logistic regression analyses)
A lower mean examination grade at A level and the late offer of a place were highly significant predictors in all students (likelihood ratio 2.19, 95% confidence interval 1.37 to 3.51, and 1.98, 1.19 to 3.30, respectively) and in UK students alone (2.36, 1.40 to3.99, and 2.25, 1.27 to 3.99). Negative comments were strongly predictive in the whole group (2.25, 1.44 to 3.50) but not in UK students alone. Men and those with lower GCSE science scores were more likely to struggle in the entire group (1.70, 1.09 to2.65, and 2.13, 1.12 to 4.05) but again not in UK students alone. Non-white ethnicity was a highly significant predictor in UK students (2.77, 1.52 to 5.05).
Validation of negative comments
Eighteen members of the interviewers' panel returned completed questionnaires (78% response rate). For the 17 comments that we (JY and DJ) thought were potentially negative, overall the interviewers agreed they were more likely to be negative. For the eight comments that we (JY and DJ) considered positive, again, the interviewers agreed they were more likely to be positive. The results of the negative comment survey are available from the corresponding author.
The review of 62 statements by an independent researcher revealed satisfactory inter-rater agreement with the identification of negative comments (κ = 0.69).