Over 13,484 anonymised patient data from 64 trusts were collected from a freedom of information act designed to examine the charging patterns from 133 trusts (figure 4).
Figure 4 – Consort diagram of data sets included in the final analysis
3.1 Nationality, country of birth and date of entry into the UK
Only nine trusts provided data sets which included information on year of entry – this amounted to 476 patients. Of these 476, no patients were known to have entered the UK before 1971, with 207 patients for whom year of entry was ‘unknown’ or the data was missing. In this group 110 were from former commonwealth countries representing 23% of the cohort who could have been someone from the Windrush generation.
9953 patients from 44 trusts were included in the analysis on pay patterns by nationality. The breakdown by nationality is shown in figure 2. Most patients invoiced were from Nigeria (1436), India (1170), USA (729), Pakistan (606), China (410), Ghana (235) and Canada (206), with a full breakdown of nationality in additional file 1. 2481 patients had data missing on their nationality. 797 patients had their nationality recorded as ‘unknown’, ‘other’ or ‘not stated’. An adjusted linear regression model showed that patients from North America are charged significantly less than those from other continents (Β= -2.4,p=0.016), f2<0.02.
3.2 Sex
The dataset has shown that patients charged were more likely to be female than male, with 63% (6,613) of over 10,000 patients being female. Figure 3a shows the boxplots of cost by sex. The multiple linear regression results show that women are charged significantly more money for health care than men, (Β=-3.129, 0.002), however the effect size was very small, f2<0.02.
Figure 3a– Box plot showing median, interquartile range and 1.5*interquartile range by sex
3.3 Age
Figures 3b-e shows the median amount charged to patients in each age group by sex. Examining the total amount charged for each age group by sex shows the large proportion of the total cost that is attributed to women of reproductive age. Nearly a third of all patients invoiced, 30.12% (3165), were women of reproductive age, and female patients were significantly more likely than men to be included in this age group (unadjusted OR 2.50 [95% CI 2.29-2.72).
Patients over 65 have significantly higher health costs (Β=2.54,0.01), and those under 16 are charged significantly less (Β= -5.513,p<0.0001), f2<0.02.
Figure 3b– Box plot showing median, interquartile range and 1.5*interquartile range by sex from children aged 0-15 years
Figure 3c– Box plot showing median, interquartile range and 1.5*interquartile range by sex from adults aged 16-40 years
Figure 3d– Box plot showing median, interquartile range and 1.5*interquartile range by sex from adults aged 41-64 years
Figure 3e– Box plot showing median, interquartile range and 1.5*interquartile range by sex from adults aged over 65
3.4 Ethnicity
Ethnicity data was provided by 12 trusts. Of these 2,872 patients had missing data, leaving 3,351 eligible for analysis (figure 5). In this group, the largest patient ethnicity was “any other ethnic group, with 1,234 patients in this category. After this, those invoiced were most likely to be of mixed-ethnicor multi-ethnic, with the largest subgroup being white/black African (856). Linear regression found no significant relationship between ethnicity and cost of health care.
Table 1 – Number of non-EEA patients of each ONS ethnicity category that were invoiced
There is no ethnicity data for tourists, those travelling for business or undocumented migrants with which to compare our data. Comparing against the number of finished admission episodes (NHS digitals coding for hospital visits) for 2017-18 compiled by NHS digital, the largest ethnic group was ‘white’, 12.9 million admissions of 16.6 million total, however ‘any other ethnic group’ had the highest rate of admissions per 100,000 of the population, and mixed ethnicity had the lowest.20
Figure 5 – Ethnicity of patients (%), for the study population, UK born population and the British non-UK born population (using census data 2011)
3.5 Urgency
The determined urgency of treatment, (additional file 1) was provided by 7 trusts. Since October 2017 this information has affected patient care as non-urgent care requires upfront payment in full before treatment commences. This provided 559 patients of which 268 were immediately necessary, 138 urgent and 153 non-urgent (figure 6). A linear regression model adjusted for age and sex showed that there was significant difference of the cost of treatment by urgency category (Β= 7.57,p<0.0001), f2=0.03.with the more urgent the treatmet the greater the cost of treatment, with a small to medium effect size.
Figure 6– Box plot showing median, interquartile range and 1.5*interquartile range for non-urgent, urgent and immediately necessary treatment