Sir, the article A ten year experience of medical emergencies at Birmingham Dental Hospital (BDJ 2018; 224: 89–91) made interesting reading but it appears the authors have got their maths and methodology completely wrong.

In the 'results' paragraph the authors explain 24 out of 119 cases were excluded from the analysis, because of missing information, leaving 95 cases. This seems to be supported by Figure 1 (bar chart with in total 95 people). Why then, in the results and summary, do they include these 24 cases again and use 119 as the denominator? You cannot exclude cases from analysis and then include them again in the denominator in your results and summary. But why are six cases with 'not enough information' included still, according to Table 1? Why were those not excluded as well? The total frequency of medical emergencies in Table 1 adds up to 115. Therefore, it appears another four were excluded (119 − 4 = 115). The authors state that in four instances there were multiple medical emergencies. So, 91 cases had one emergency only, leaving four cases with on average six simultaneous medical emergencies each (91 + 24 = 115)? That would have been very bad luck for these four individuals!

Even allowing for the basic error (re-including excluded cases) the figures in the results and summary are wrong. The authors are giving the impression 119 is the denominator. Cross referencing with the data in Table 1, it is hard to follow how the authors come to the percentages stated in the summary unless, at times, they use 115 as the denominator but not all the time.

For example, asthmatic attack occurred three times. The authors state this is 2.6%. As a percentage of 95 that is 3.2%, as a percentage of 119 this would be 2.5%, as a percentage of 115 this would be 2.6%. Vasovagal syncope occurred (according to Table 1) 42 times. If divided by 119 this would be 35%, not 36.5%. It appears 115 was used as the denominator here (42 being 36.5% of 115). They continue: 'cardiac arrest, stroke and iatrogenic events 1.7%'. Do they mean 1.7% each? Myocardial infarction occurred once (1/119) × 100% = 0.84%. Unless the authors used 115 again as denominator, then it would make sense (0.86% rounded up to 0.9%). But then why use 119 as the denominator in 'paramedic attendance' (48/119) × 100% = 40.3% as stated?

The authors conclude that BDH has a robust emergency protocol and recordkeeping system. How can a recordkeeping system with an incomplete record rate of 20% (24/119) ever be called robust?

The authors, S. Sooch, A. Kaur and B. Ahmed respond: We thank your reader for their comments. We need to acknowledge that the Datix system is a web-based online incidence reporting software usually completed by the dentist but in some cases by another member of the dental team. As investigators we access these data from the spreadsheet held by the governance manager. To our knowledge, this is the only way we can capture medical emergencies as reported by the dental team for our hospital. The Datix online reporting involves free text data input by the person filling in the report. This explains some of the discrepancies as some of the data are not documented and therefore we need to account for this. In saying this, we were stating that the nature of the medical emergency was known in 119 cases, however any additional demographic information (ie age) had been missing from some of the data entries; therefore the bar graph depicted a total of 95. Hence why Figure 1, which showed age groups of people affected and Table 1, which shows the frequency of medical emergencies, add up to different figures, as we were trying to depict these separately.

We excluded the four cases from 119 to give us a denominator of 115 (not from 95 to 91), as the dataset again was incomplete and we decided to continue with the 115 definitive cases and outcomes available.

We used the denominator 119 for paramedics called to the hospital, as the data were always available for each case regarding paramedic attendance as demonstrated in Figure 2.

Patient safety is a core value of our hospital; we like many other hospitals use the Datix system as a platform in recording untoward events such as medical emergencies. We accept the Datix system could be improved as it relies heavily on the information input by the reporters: missing data, a wrong clinical diagnosis input in the system can give rise to discrepancies; however, changing the system was beyond the scope of our project.