J R Soc Med OnlineFirst, published on July 7, 2015 as doi:10.1177/0141076815588321
Research Journal of the Royal Society of Medicine; 0(0) 1–12 DOI: 10.1177/0141076815588321
Ethnic variations in morbidity and mortality from lower respiratory tract infections: a retrospective cohort study Colin R Simpson1, Markus FC Steiner2, Genevieve Cezard1, Narinder Bansal3, Colin Fischbacher4, Anne Douglas1, Raj Bhopal1, Aziz Sheikh1,5 on behalf of the SHELS researchers 1
Edinburgh Ethnicity and Health Research Group, Centre for Population Health Sciences, The University of Edinburgh, Edinburgh, UK 2 Department of Child Health, School of Medicine, University of Aberdeen, Aberdeen, UK 3 Cardiovascular Epidemiology Unit, The Department of Public Health and Primary Care, University of Cambridge, Cambridge, UK 4 NHS National Services, Edinburgh, UK 5 Division of General Internal Medicine and Primary Care, Brigham and Women’s Hospital/Harvard Medical School, Boston MA, USA Corresponding author: Colin R Simpson. Email: [email protected]
Abstract Objective: There is evidence of substantial ethnic variations in asthma morbidity and the risk of hospitalisation, but the picture in relation to lower respiratory tract infections is unclear. We carried out an observational study to identify ethnic group differences for lower respiratory tract infections. Design: A retrospective, cohort study. Setting: Scotland. Participants: 4.65 million people on whom information was available from the 2001 census, followed from May 2001 to April 2010. Main outcome measures: Hospitalisations and deaths (any time following first hospitalisation) from lower respiratory tract infections, adjusted risk ratios and hazard ratios by ethnicity and sex were calculated. We multiplied ratios and confidence intervals by 100, so the reference Scottish White population’s risk ratio and hazard ratio was 100. Results: Among men, adjusted risk ratios for lower respiratory tract infection hospitalisation were lower in Other White British (80, 95% confidence interval 73–86) and Chinese (69, 95% confidence interval 56–84) populations and higher in Pakistani groups (152, 95% confidence interval 136–169). In women, results were mostly similar to those in men (e.g. Chinese 68, 95% confidence interval 56–82), although higher adjusted risk ratios were found among women of the Other South Asians group (145, 95% confidence interval 120–175). Survival (adjusted hazard ratio) following lower respiratory tract infection for Pakistani men (54, 95% confidence interval 39–74) and women (31, 95% confidence interval 18–53) was better than the reference population. Conclusions: Substantial differences in the rates of lower respiratory tract infections amongst different ethnic groups in Scotland were found. Pakistani men and women had particularly high rates of lower respiratory tract infection
hospitalisation. The reasons behind the high rates of lower respiratory tract infection in the Pakistani community are now required.
Keywords Respiratory tract infections, secondary care, death, ethnic groups, incidence, hospital readmission, hospitalisation
Introduction Acute lower respiratory infections are associated with significant morbidity and mortality and are one of the leading causes of years of life lost due to premature mortality.1 Lower respiratory tract infection is an umbrella term and includes (among other diseases) influenza and pneumonia. In the UK, disease rates (in particular pneumonia and influenza) usually peak during the winter and spring months (i.e. September to February and March to May).2 Risk factors that increase the likelihood of acute lower respiratory tract infections include chronic respiratory diseases (e.g. asthma and chronic obstructive pulmonary disease), tobacco smoking and immune suppression.3 Of the lower respiratory tract infections, pneumonia causes the most deaths, with four million each year worldwide. Pneumonia is particularly prevalent in children and the elderly.4–6 For lower respiratory tract infections, reduction in modifiable risk factors, e.g. smoking and increased uptake of vaccination,7–9 represents an important approach to reduce illness and death. Variations in lower respiratory tract infections by ethnic group have been noted previously in the United States (US), New Zealand and Australia.10–15 These
! The Royal Society of Medicine 2015 Reprints and permissions: sagepub.co.uk/journalsPermissions.nav Downloaded from jrs.sagepub.com by guest on November 11, 2015
2
Journal of the Royal Society of Medicine 0(0) Figure 1. Overview of Record Linkage Process (CHI and census numbers were encrypted).
Census Database
Health Database CHI Number
Record Linkage Personal Identifiers Personal Identifiers
Census Number
CHI Number Census Number (Look-up Table)
Health Information
studies found disparities in the incidence of lower respiratory tract infections among American Indian and Alaska native children and pneumonia among Black US adults and Maori in New Zealand. Some of these studies, however, did not adjust for confounding socioeconomic factors.11,13,14 National studies are needed to explore heterogeneity within broad ethnic groups, to maintain valid surveillance of trends and inequalities in diseases such as pneumonia, to set priorities (e.g. vaccination strategies and smoking cessation) and to ensure equitable service delivery.16 Furthermore, ethnic variations in infectious respiratory diseases need periodic documentation. This is because disease patterns are altering over time as a result of changing environments and risk factor patterns (e.g. pollutants, housing, diet and smoking), particularly following migration.17 Also lower respiratory tract infection rates may relate to travel, specifically religious travel such as the Hajj in which many people congregate at the same site and outbreaks of airborne/droplet borne infections (e.g. influenza, Middle East Respiratory Syndrome and meningitis) can occur.18 Thus far, there have been few studies investigating ethnic variations in hospitalisations and deaths resulting from lower respiratory tract infections. Building on related work,19 we undertook a national observational study, adjusting for socioeconomic confounding to identify ethnic group differences for lower respiratory tract infections.
Methods We linked records from Scotland’s routinely collected hospital discharge records (SMR01) and the 2001 Census using probabilistic linkage methods (using information from multiple variables to maximise the
Ethnicity Information
chances of correctly linking disparate information on individuals) creating a retrospective cohort study of 4.65 million individuals; this represented over 90% of the Scottish population in 2001 (Figure 1). All personal identifiers including the encrypted Community Health Index identifier were removed from the data before analysis began. Over 80% linkage was achieved for each ethnic group.19
Ethnic groups The ethnic group categories were those of the 2001 Census (Box 1) as reported by the head of household or individuals. These categories include the largest minority ethnic groups in Scotland in 2001. Where possible, aggregation was minimised to avoid missing important group differences, but for outcomes (discussed below) for which there were small numbers, we combined Bangladeshis with Other South Asians and the Caribbean, African and Black Scottish or Other Black as African origin to minimise the risk of inadvertently breaching anonymity.
Demographic and socioeconomic factors In addition to age and sex, sociodemographic factors extracted from census data included: highest educational qualification of the individual (and separately the household), occupation, household tenure and quintile of Scottish Index of Multiple Deprivation.20 The Scottish Index of Multiple Deprivation is the Scottish Government’s official measure of area-based multiple deprivation. It is based on 31 indicators in six individual domains of current income, employment, housing (although not household crowding), health, education, skills
Downloaded from jrs.sagepub.com by guest on November 11, 2015
Simpson et al.
3
and training and geographic access to services and telecommunications.
data. Events were identified in all age groups from any of the diagnoses codes recorded on discharge from hospital or from the death certificates. A first admission was defined as a hospital admission without prior hospital admission for lower respiratory tract infection, time to death from first admission to death (any cause) and time for first admission to a second readmission were identified.
Outcomes of interest Hospital admissions and deaths for lower respiratory tract infection, pneumonia and influenza (after hospitalisation for any of these diseases) were collected for a nine-year period from May 2001 to April 2010 (Box 2). These data (SMR01) have been found to be reliable, with completeness and accuracy rates exceeding 90%.21 Their value for epidemiologic research has been repeatedly demonstrated.22 The time period chosen reflects the time between the 2001 census and the most recent available outcomes
Statistical analysis We calculated age-adjusted rates and risk ratios (adjusted risk ratio) per 100,000/year with a personyears-at-risk denominator, using Poisson regression models with robust variance. As differences between sexes were found in previous analyses,16,19 an a priori decision was taken to stratify our analysis by sex. The White Scottish population was used as the reference population (risk ratio ¼ 100) and all comparisons relate to the reference as stated or implied. We identified first events over a nine-year period: i.e. May 2001 to April 2010. No test for interaction was carried out because of the lack of statistical power, so observations on potential effect modification are interpreted as hypothesis generating and preliminary. We further adjusted risk ratios for country of birth (born in the UK or born outside the UK) and Scottish Index of Multiple Deprivation quintile as a proxy for socioeconomic status. We report overall age-adjusted survival to a cause-specific death following a first hospital admission for the same disease, for outcomes where there were sufficient numbers to allow analyses and satisfy disclosure control protocols. Survival analysis was carried out using Cox regression to calculate hazard ratios adjusted for age (adjusted hazard ratio) using the White Scottish population as reference. Survival to death and to readmission covered the nine-year follow-up period. Proportional hazards assumption was tested with the graphical method using log-log (-ln{-ln(survival)}) plots. Analyses were undertaken using the statistical
Box 1. Scotland 2001 Census – self-reported ethnic groups. White Scottish Other White British White Irish Other White Any mixed background Indian Pakistani Bangladeshi Other South Asian Caribbean African Black Scottish or Other Black Chinese Other ethnic group
Box 2. International Classification of Disease (ICD)-9 and ICD-10 codes for respiratory conditions of interest. ICD10 title
ICD9
ICD10
Lower respiratory infections (influenza, pneumonia, acute bronchitis and bronchiolitis, respiratory tuberculosis lung abscess and pleural empyema)
480–485, 486, 487, 514, 466.0, 519.8, 466.1, 513, 510, 010–012
J10–J18, J20–J22, A15–A16, J85–J86
Pneumonia
480–486 and 514
J12–J18
Influenza
487
J10–11
Downloaded from jrs.sagepub.com by guest on November 11, 2015
4
Journal of the Royal Society of Medicine 0(0)
analysis package SAS v9.3 (SAS Institute Inc., Cary, NC, USA).
Ethical approval Ethical approval was granted by the Scotland A Research Ethics Committee. Permissions for data linkage and access to the Community Health Index were obtained from the Privacy Advisory Committee of the NHS Information Services Division, the General Register Office for Scotland (now National Records Scotland) and the Community Health Index Advisory Group. Access to data, analyses and release of outputs followed a prespecified protocol to ensure confidentiality and to prevent disclosure of any potentially identifying data.
adjusted risk ratios (145; 95% confidence interval 120–175) were found among Other South Asian women (Table 2). Also (as with White Irish men), after adjustment, risk ratios among White Irish women became significantly lower than the reference population (from 103; 95% confidence interval 95– 112 to 92; 95% confidence interval 85–100). Survival for women was better in minority ethnic groups (although this was nonsignificant for Indian women). Compared to the reference population, readmission rates after a first hospitalisation for lower respiratory tract infection were significantly lower among other White British men and women and higher among Indian and Pakistani women (Table 3).
Pneumonia Results During this retrospective study with a nine-year follow-up, there were 38 million person-years at risk available for our analyses (Table 1). The majority of person-years at risk were for the White Scottish population. Lower respiratory tract infection accounted for 201,873 hospitalisations (a rate of 528 per 100,000 person-years at risk). Pneumonia accounted for 124,764 hospitalisations (324 per 100,000 person-years at risk) and influenza 2136 hospitalisations (6 per 100,000 person-years at risk). There were 51,533 deaths from lower respiratory tract infection (134 per 100,000 person years at risk); 39,780 of deaths were due to pneumonia (104 per 100,000 person years at risk).
Lower respiratory tract infection White Scottish men had a rate of lower respiratory tract infection of 541 per 100,000 patient-years. When compared with this reference population, the adjusted risk ratios for lower respiratory tract infection hospitalisation were significantly lower among Other White British and Chinese groups (Table 1) and significantly higher in Indian and Pakistani groups. In White Irish men, after adjustment for age, socioeconomic status and country of birth, the adjusted risk ratios attenuated from a significantly higher risk ratio to almost the reference value (i.e. from an unadjusted risk ratio: 117; 95% confidence interval 107–129 to an adjusted risk ratio of 101; 95% confidence interval 93–110). Survival in men following lower respiratory tract infection was mostly similar or better in minority ethnic groups, in particular Indian and Pakistani groups. In women, the results (i.e. in comparison with White Scottish women) were similar to those in men, although significantly higher
Among men, the adjusted risk ratios for pneumonia hospitalisation were low in Other White British and Chinese, who had the lowest adjusted risk ratio of pneumonia hospitalisations of any group (62; 95% confidence interval 45–84; Table 4). High adjusted risk ratios were found in the Any Mixed Background and Pakistani groups. As with lower respiratory tract infections, after adjustment, the risk ratios for pneumonia attenuated almost to the reference value in the White Irish men. Survival following pneumonia was mostly similar or better in men of minority ethnic groups. In women, significantly higher adjusted risk ratios were found among Other South Asian (148; 95% confidence interval 115–191) and lower adjusted risk ratios (90; 95% confidence interval 81–100) in Other White (Table 5). As in men, survival for women was similar or better in minority ethnic groups. Similar risks of readmission rates to lower respiratory tract infections, after a first hospitalisation for pneumonia were found (Table 3).
Influenza Influenza hospitalisation adjusted risk ratios were found to be higher in Pakistani men (235; 95% confidence interval 134, 412) than the reference group. This was the only significant group difference (in men and women) for influenza (Supplemental Table 1).
Discussion Differences in the risks of hospital admission were found for lower respiratory tract infections including pneumonia and influenza, in particular with higher risks for Pakistani men and substantially lower risks in Chinese populations. Risks of readmission were higher for Indian and Pakistani women.
Downloaded from jrs.sagepub.com by guest on November 11, 2015
Downloaded from jrs.sagepub.com by guest on November 11, 2015
1254
977
120
168
406
81
87
94
43
White Scottish
Other White British
White Irish
Other White
Any mixed background
Indian
Pakistani
Other South Asian
African Origin
Chinese
All other ethnic group
PY at risk
25,771
52,884
24,955
27,365
110,071
50,677
43,123
223,523
159,151
12,436,76
16,312,393
N
381.9
340.9
684.2
593.0
803.1
557.2
669.8
475.5
633.8
422.7
540.6
Poisson ratesa (for 100,000 PY)
70.6 (50.5, 98.9)
63.1 (46.2, 86.0)
126.6 (94.0, 170.4)
109.7 (90.0, 133.6)
148.5 (133.9, 164.8)
103.1 (88.1, 120.5)
123.9 (103.6, 148.2)
87.9 (77.9, 99.3)
117.2 (106.9, 128.5)
78.2 (70.0, 87.3)
100.0 (–)
RR and 95% CI
Poissonb relative risks Age adjusted
70.6 (54.6, 91.2)
65.8 (53.1, 81.5)
118.0 (97.6, 142.6)
107.9 (86.9, 134.0)
147.7 (132.6, 164.5)
112.6 (96.7, 131.0)
120.9 (100.7, 145.3)
90.3 (82.6, 98.7)
107.7 (99.4, 116.7)
84.5 (78.0, 91.4)
100.0 (–)
RR and 95% CI
Age and SIMD adjustedb
Note: Poisson rates and relative risk and hazard ratio by ethnic group for men. PY, person years; RR, relative risk; SIMD, Scottish index for multiple deprivation; COB, country of birth; HR, hazard ratio. a Poisson rates are derived from relatives risks with age adjustment using the White Scottish rate as the reference. b Calculated relative to the White Scottish population at 2001 census. c Over a nine-year period between 1st May 2001 and 30th April 2010.
88,191
5912
N
Ethnic group
All lower respiratory infection
Table 1. All lower respiratory tract infection hospitalisations and deaths for men.
74.8 (52.4, 106.8)
66.7 (51.2, 87.0)
129.8 (98.2, 171.6)
113.3 (92.6, 138.7)
154.1 (140.0, 169.5)
108.7 (95.0, 124.5)
118.3 (97.0, 144.1)
92.3 (81.4, 104.6)
107.2 (97.1, 118.2)
71.9 (64.5, 80.1)
100.0 (–)
RR and 95% CI
Age and COB adjustedb
73.6 (55.1, 98.2)
68.6 (55.8, 84.4)
120.1 (97.9, 147.2)
110.3 (90.0, 135.0)
151.8 (136.1, 169.4)
117.1 (101.3, 135.3)
116.8 (96.8, 141.0)
93.6 (85.4, 102.6)
100.9 (92.5, 110.1)
79.5 (73.2, 86.3)
100.0 (–)
RR and 95% CI
Age, SIMD and COB adjustedb
7 81.8 (39.0, 171.6)
18 88.2 (55.6, 140.1)
6 47.5 (21.3, 105.8)
9 56.6 (29.4, 108.8)
36 53.6 (38.7, 74.4)
20 64.2 (41.4, 99.5)
21 101.6, (66.2, 155.8)
283 93.9 (83.5, 105.6)
353 99.3 (89.4, 110.3)
1594 91.4 (86.9, 96.2)
22651 100.0 (–)
HR and 95% CIc
Survival to death from LRTI
Simpson et al. 5
Downloaded from jrs.sagepub.com by guest on November 11, 2015
70
87
51
African origin
Chinese
All other ethnic group
29,185
52,777
22,273
22,301
108,226
46,110
46,518
259,738
178,557
13,39,624
1,786,9735
PY at risk
403.9
338.2
654.7
788.8
805.7
556.3
566.1
412.7
552.9
436.2
534.5
75.6 (60.7, 94.0)
63.3 (51.1, 78.4)
122.5 (101.4, 148.0)
147.6 (122.1, 178.3)
150.7 (127.6, 178.1)
104.1 (94.4, 114.8)
105.9 (87.1, 128.7)
77.2 (70.1, 85.0)
103.4 (95.3, 112.3)
81.6 (75.6, 88.1)
100.0 (–)
RR and 95% CI
Poisson relative risksb Age adjusted
77.5 (61.0, 98.4)
66.6 (55.4, 79.9)
116.7 (97.1, 140.3)
144.0 (119.8, 173.0)
148.6 (130.7, 169.0)
112.1 (96.6, 130.1)
102.0 (80.9, 128.7)
80.4 (73.7, 87.6)
94.8 (87.9, 102.3)
87.3 (81.4, 93.6)
100.0 (–)
RR and 95% CI
Age and SIMD adjustedb
79.0 (58.8, 106.1)
66.2 (54.2, 80.9)
123.6 (99.0, 154.3)
149.1 (120.6, 184.3)
153.7 (128.4, 184.2)
107.4 (94.1, 122.7)
103.0 (82.0, 129.4)
80.3 (72.8, 88.6)
97.5 (88.7, 107.2)
77.3 (71.0, 84.1)
100.0 (–)
RR and 95% CI
Age and COB adjustedb
79.3 (61.8, 107.7)
68.1 (56.4, 82.3)
117.3 (96.4, 142.7)
144.7 (119.9, 174.7)
150.2 (131.8, 171.2)
113.9 (97.7, 132.9)
100.5 (81.1, 124.5)
82.0 (75.4, 89.1)
92.0 (84.8, 99.8)
84.9 (78.9, 91.4)
100.0 (–)
RR and 95% CI
Age, SIMD and COB adjustedb
Note: Poisson rates and relative risk and hazard ratio by ethnic group for women. PY, person years; RR, relative risk; SIMD, Scottish index for multiple deprivation; COB, country of birth; HR, hazard ratio; LRTI, lower respiratory tract infection. a Poisson rates are derived from Relatives Risks with age adjustment using the White Scottish rate as the reference. b Calculated relative to the White Scottish population at 2001 census. c Over a nine-year period between 1 May 2001 and 30 April 2010. d Results suppressed due to small numbers.
85
322
Pakistani
Other South Asian
122
858
Other White
Indian
1282
White Irish
116
6031
Other White British
Any mixed background
95516
White Scottish
Ethnic group
All lower respiratory event
Poisson ratesa (for 100,000 PY)
Table 2. All lower respiratory tract infection hospitalisations and deaths for women.
d
8
11
13
13
18
15
172
356
1491
24,458
All
41.5 (20.8, 83.1)
110.3 (61.1, 199.3)
98.7 (57.3, 170.0)
30.6 (17.8, 52.7)
83.7 (52.7, 132.9)
55.5 (33.5, 92.1)
74.4 (64.0, 86.4)
92.3 (83.2, 102.5)
88.0 (83.5, 92.7)
100.0 (–)
HR and deathsc 95% CI
Survival to death from LRTI
6 Journal of the Royal Society of Medicine 0(0)
Downloaded from jrs.sagepub.com by guest on November 11, 2015
239
172
18
26
74
13
7
16
9
White Irish
Other White
Any mixed background
Indian
Pakistani
Other South Asian
African Origin
Chinese
All other ethnic group
120.6 (62.8, 231.9)
79.8 (46.3, 137.4)
59.5 (28.4, 124.8)
85.2 (48.4, 150.1)
103.3 (80.9, 131.8)
87.2 (57.4, 132.6)
95.0 (58.2, 155.1)
95.4 (81.9, 111.2)
104.5 (91.6, 119.3)
82.5 (77.1, 88.3)
100.0 (–)
HR and 95% CI
Overall readmission
a
6
a
7
22
13
10
69
85
368
5809
Pneumonia
N of second event
a
91.7 (41.2, 204.3)
a
152.5 (72.6, 320.2)
103.0 (67.7, 156.6)
147.5 (85.6, 254.4)
116.7 (60.7, 224.6)
103.1 (81.0, 131.2)
102.6 (82.5, 127.6)
85.5 (76.7, 95.2)
100.0 (–)
HR and 95% CI
Overall readmission
Note: LRTI, lower respiratory tract infection; HR, hazard ratio; CI, confidence interval. a Results suppressed due to small numbers.
937
15409
White Scottish
Other White British
LRTI
Male ethnic group
N of second event
All Other Ethnic Group
Chinese
African Origin
Other South Asian
Pakistani
Indian
Any mixed background
Other White
White Irish
Other White British
White Scottish
Female ethnic group
Table 3. Risk of readmission for any diagnosis following first hospital admission for LRTI and pneumonia.
8
12
11
13
79
34
16
152
239
862
15633
LRTI
N of second event
97.6 (46.5, 204.9)
77.7 (43.0, 140.4)
103.0 (55.4, 191.5)
119.9 (68.0, 211.1)
159.6 (126.5, 201.3)
163.8 (113.1, 237.4)
83.0 (50.8, 135.6)
98.5 (83.7, 115.9)
112.0 (98.2, 127.7)
80.4 (74.9, 86.2)
100.0 (–)
HR and 95% CI
Overall readmission
a
a
a
a
a
a
158.9 (101.2, 249.6)
186.2 (100.1, 346.4)
a
94.3 (71.0, 125.3)
117.0 (93.4, 146.6)
72.2 (63.7, 81.7)
100.0 (–)
HR and 95% CI
Overall readmission
a
a
19
10
a
48
80
267
5333
Pneumonia
N of second event
Simpson et al. 7
635
Other White
Downloaded from jrs.sagepub.com by guest on November 11, 2015
40
46
23
African Origin
Chinese
All other ethnic group
25,843
53,089
25,116
27,527
110,919
51,006
43,255
224,653
160,540
12,50,532
16,420,136
PY at risk
245.9
193.6
363.1
368.1
424.8
320.4
504.3
328.2
428.2
290.0
339.5
72.4 (49.5, 106.1)
57.0 (34.3, 94.7)
107.0 (72.6, 157.5)
108.4 (83.2, 141.3)
125.1 (108.8, 143.8)
94.4 (76.5, 116.4)
148.5 (122.7, 179.9)
96.7 (82.6, 113.1)
126.1 (111.9, 142.1)
85.4 (76.0, 96.0)
100.0 (–)
RR and 95% CI
Poisson relative risksb Age adjusted
72.5 (50.8, 103.6)
59.1 (43.5, 80.3)
99.7 (77.5, 128.3)
106.5 (80.9, 140.3)
125.0 (109.1, 143.1)
102.9 (87.2, 121.4)
145.0 (115.7, 181.6)
99.2 (88.7, 111.0)
115.7 (104.5, 128.1)
92.4 (84.3, 101.3)
100.0 (–)
RR and 95% CI
Age and SIMD adjustedb
76.4 (47.9, 121.7)
60.4 (39.7, 92.0)
109.6 (72.5, 165.6)
111.8 (88.8, 140.7)
130.4 (114.1, 149.0)
99.8 (81.7, 122.1)
141.7 (116.5, 172.3)
101.4 (87.3, 117.8)
115.3 (101.6, 130.9)
78.5 (69.6, 88.6)
100.0 (–)
RR and 95% CI
Age and COB adjustedb
Note: Poisson rates and relative risk and hazard ratio by ethnic group for men. PY, person years; RR, relative risk; SIMD, Scottish index for multiple deprivation; COB, country of birth; HR, hazard ratio. a Poisson rates are derived from relatives risks with age adjustment using the White Scottish rate as the reference. b Calculated relative to the White Scottish population at 2001 census. c Over a nine-year period between 1 May 2001 and 30 April 2010. d Results suppressed due to small numbers.
43
184
Other South Asian
Pakistani
85
809
White Irish
Indian
3818
Other White British
76
55,751
White Scottish
Any mixed background
Pneumonia events
Ethnic group
Poisson ratesa (per 100,000 PY)
Table 4. Pneumonia hospitalisations and deaths for men.
75.4 (50.8, 111.9)
61.6 (45.4, 83.7)
101.3 (76.7, 133.8)
108.6 (84.2, 140.0)
128.8 (111.2, 149.2)
107.1 (90.6, 126.5)
140.1 (111.9, 175.3)
102.7 (91.7, 115.1)
108.6 (97.3, 121.3)
87.1 (79.0, 96.0)
100.0 (–)
RR and 95% CI
Age, SIMD and COB adjustedb
7
16
d
d
27
14
18
212
277
1231
17776
All
103.5 (49.4, 217.2)
88.2 (54.0, 144.1)
65.0 (44.6, 94.8)
66.1 (39.1, 111.6)
106.3 (67.0, 168.8)
89.8 (78.5, 102.9)
99.6 (88.4, 112.1)
92.0 (86.8, 97.5)
100.0 (–)
HR and deathsc 95% CI
Survival to death from pneumonia
8 Journal of the Royal Society of Medicine 0(0)
504
Other White
Downloaded from jrs.sagepub.com by guest on November 11, 2015
33
45
22
African Origin
Chinese
All other ethnic group
29,308
52,882
22,464
22,429
109,070
46,363
46,745
260,933
180,068
13,47,169
17,99,4602
PY at risk
216.5
211.7
367.8
484.9
375.9
307.1
378.0
269.5
368.1
287.1
321.8
67.3 (51.0, 88.8)
65.8 (51.7, 83.6)
114.3 (91.7, 142.5)
150.7 (109.0, 208.3)
116.8 (92.5, 147.7)
95.4 (78.6, 115.8)
117.5 (90.0, 153.4)
83.8 (74.7, 93.9)
114.4 (103.1, 126.9)
89.2 (81.3, 97.9)
100.0 (–)
RR and 95% CI
Poisson relative risksb Age adjusted
69.0 (50.7, 93.9)
69.0 (55.8, 85.2)
109.5 (83.7, 143.1)
147.3 (110.9, 195.6)
115.6 (95.8, 139.3)
102.3 (81.8, 128.0)
112.9 (85.0, 150.0)
87.0 (78.3, 96.6)
105.0 (94.9, 116.1)
95.3 (87.7, 103.7)
100.0 (–)
RR and 95% CI
Age and SIMD adjustedb
71.0 (50.0, 100.7)
69.8 (52.5, 92.9)
115.2 (86.6, 153.4)
152.1 (116.2, 199.1)
120.5 (95.4, 152.4)
99.9 (80.5, 124.0)
113.1 (84.1, 152.3)
88.1 (78.4, 98.9)
106.1 (93.9, 119.8)
83.3 (74.7, 92.7)
100.0 (–)
RR and 95% CI
Age and COB adjustedb
Note: Poisson rates and relative risk and hazard ratio by ethnic group for women. PY, person years; RR, relative risk; SIMD, Scottish index for multiple deprivation; COB, country of birth; HR, hazard ratio. a Poisson rates are derived from relatives risks with age adjustment using the White Scottish rate as the reference. b Calculated relative to the White Scottish population at 2001 census. c Over a nine-year period between 1 May 2001 and 30 April 2010. d Results suppressed due to small numbers.
44
119
Other South Asian
Pakistani
56
805
White Irish
Indian
3659
Other White British
65
57,902
White Scottish
Any mixed background
Pneumonia event
Ethnic group
Poisson ratesa (for 100,000 PY)
Table 5. Pneumonia hospitalisations and deaths for women.
71.2 (50.8, 99.9)
71.5 (56.3, 90.9)
110.0 (84.1, 143.9)
148.1 (114.5, 191.4)
117.8 (96.7, 143.6)
105.2 (82.7, 133.9)
110.3 (83.9, 145.0)
89.7 (80.6, 99.8)
100.3 (90.0, 111.7)
91.4 (83.5, 100.1)
100.0 (–)
RR and 95% CI
Age, SIMD and COB adjustedb
d
7
8
11
9
13
13
124
258
1146
18,703
All
51.0 (24.3, 107.0)
101.7 (50.9, 203.4)
128.2 (71.0, 231.5)
38.5 (20.0, 74.1)
97.4 (56.6, 167.9)
57.7 (33.5, 99.3)
75.7 (63.4, 90.3)
88.9 (78.6, 100.5)
89.9 (84.6, 95.4)
100.0 (–)
HR and deathsc 95% CI
Survival to death from pneumonia
Simpson et al. 9
10
Journal of the Royal Society of Medicine 0(0)
Comparison of findings with other published work The literature on ethnicity and lower respiratory tract infections, particularly among European populations, is sparse – however, risks were found to be higher in indigenous Aboriginal children in Australia10 and Native Americans in the US.11 Comparatively higher rates of community-acquired pneumonia have been found among Black adults in the US12 and among Maori in New Zealand.13 For influenza, indigenous peoples also have high rates of disease and poorer outcomes.14 The differences, however, in categorisation of race/ethnicity and system of healthcare in the US, Australia and New Zealand prevent meaningful comparisons with our data. In the UK, no difference in the rate of influenza between ethnic groups has been previously noted. However, a study by Nyland et al. set out to determine whether care pathway differences existed between non-White and White patients admitted with laboratory-confirmed pandemic influenza A (H1N1). They found that the non-White group was less likely to receive preadmission antibiotics.15 However, there were no significant differences in delayed admission, severity at presentation for admission or likelihood of severe outcome.
Main strengths and limitations of this work The strengths of Scottish Health and Ethnicity Linkage Study include the analysis at national level, the cohort design, the availability of ethnic group and sociodemographic data (including socioeconomic status) and the analyses of many ethnic groups.23 We used Scottish Index of Multiple Deprivation because hospitalisation/death from respiratory diseases is commonest in older populations and Scottish Index of Multiple Deprivation data from the Census were available for all ages. Unlike Scottish Index of Multiple Deprivation, other alternative measures such as educational and economic status are only collected from 18 to 74 years of age. We used all diagnostic data on the hospitalisation and on mortality records (so-called multi-cause coding) and not just first hospital diagnosis and underlying cause of death. Both approaches have merits and limitations. Since we used the data primarily to compare ethnic groups, we think our approach is appropriate, especially as it maximises the number of events. Limitations include: incomplete linkage, few events in some groups (particularly for influenza) reflected in wide confidence intervals, lack of community and primary care data (where the majority of less severe lower respiratory tract infections are likely to present) and lack of linkable risk factor data, e.g. smoking (although a feasibility study to link primary care
data to the Scottish Health and Ethnicity Linkage Study database is underway). There are other important missing data including key factors along the causal pathway of disease, e.g. social and genetic determinants, disease prevention, healthcare-seeking behaviour and healthcare access (e.g. higher risks could reflect an underestimated denominator or high levels of referral to hospital in some ethnic groups), potential differential treatment of people from different ethnic backgrounds by healthcare institutions (including related to sociocultural issues or difficulties with communication) and differences between use of alternative medicine and home care in different groups (including the care of people who are dying at home). Changing disease patterns (i.e. the 2009 H1N1 pandemic influenza outbreak) and specific co-morbidities such as diabetes, tuberculosis and immunosuppression, which may be high in specific ethnic groups and contribute to lower respiratory tract infection risk, were also missing. We did not have access to data to determine changing migration over time (although we had access to country of birth from the census); we were therefore unable to adjust for the length of residence in Scotland. This is likely to have an impact, as those living for longer periods in Scotland are likely to have markedly different health and socioeconomic status than more recent migrants who had being living in Scotland for a shorter period of time. Furthermore, lower risks could reflect unrecorded emigration whereby the numerator is underestimated and the denominator overestimated. As linkage was to the 2001 census, we were unable to link data to those born in Scotland after 2001. The number of children was therefore too small to allow a separate analysis. As causative lower respiratory tract infection and pneumonia organisms are different in adults and children (requiring different control measures), a stratified analysis by age should be a priority for future research. Bias may have been introduced as some groups with complex health needs, e.g. asylum seekers and the homeless, were under-represented in the 2001 census. Linkage rates were slightly lower among ethnic minorities than among the White Scottish population, increasing the risk of bias. However, there was no appreciable variation between groups in linkage rates. The 80% linkage target achieved for all ethnic groups was determined by the investigators to be both a realistic and yet valid and searching target.
Meaning of the study results: possible mechanisms and implications for clinicians and policy makers We have shown substantial differences between ethnic groups, with low risks in Chinese and surprisingly high risks of hospitalisation (but better rates of
Downloaded from jrs.sagepub.com by guest on November 11, 2015
Simpson et al.
11
survival) among Pakistani men. This finding of higher rates of a lower respiratory tract infection hospitalisation in Pakistani men is surprising. The reference White Scottish population, with a higher prevalence of smoking when compared with other ethnic groups, should be at a higher risk of respiratory diseases.24 Vaccination, in particular with the pneumococcal polysaccharide and conjugate vaccines, haemophilus influenzae (Hib) vaccine and inactivated and live attenuated influenza vaccine, is a key preventative measure for reducing the burden of pneumonia. Although few data on specific vaccine uptake among different ethnic groups living in the UK exist, it is known that areas where more non-White groups live have lower uptake of pneumococcal polysaccharide vaccine vaccination (areas with population 99–100% White: 7.8% vs. areas with population
Research Journal of the Royal Society of Medicine; 0(0) 1–12 DOI: 10.1177/0141076815588321
Ethnic variations in morbidity and mortality from lower respiratory tract infections: a retrospective cohort study Colin R Simpson1, Markus FC Steiner2, Genevieve Cezard1, Narinder Bansal3, Colin Fischbacher4, Anne Douglas1, Raj Bhopal1, Aziz Sheikh1,5 on behalf of the SHELS researchers 1
Edinburgh Ethnicity and Health Research Group, Centre for Population Health Sciences, The University of Edinburgh, Edinburgh, UK 2 Department of Child Health, School of Medicine, University of Aberdeen, Aberdeen, UK 3 Cardiovascular Epidemiology Unit, The Department of Public Health and Primary Care, University of Cambridge, Cambridge, UK 4 NHS National Services, Edinburgh, UK 5 Division of General Internal Medicine and Primary Care, Brigham and Women’s Hospital/Harvard Medical School, Boston MA, USA Corresponding author: Colin R Simpson. Email: [email protected]
Abstract Objective: There is evidence of substantial ethnic variations in asthma morbidity and the risk of hospitalisation, but the picture in relation to lower respiratory tract infections is unclear. We carried out an observational study to identify ethnic group differences for lower respiratory tract infections. Design: A retrospective, cohort study. Setting: Scotland. Participants: 4.65 million people on whom information was available from the 2001 census, followed from May 2001 to April 2010. Main outcome measures: Hospitalisations and deaths (any time following first hospitalisation) from lower respiratory tract infections, adjusted risk ratios and hazard ratios by ethnicity and sex were calculated. We multiplied ratios and confidence intervals by 100, so the reference Scottish White population’s risk ratio and hazard ratio was 100. Results: Among men, adjusted risk ratios for lower respiratory tract infection hospitalisation were lower in Other White British (80, 95% confidence interval 73–86) and Chinese (69, 95% confidence interval 56–84) populations and higher in Pakistani groups (152, 95% confidence interval 136–169). In women, results were mostly similar to those in men (e.g. Chinese 68, 95% confidence interval 56–82), although higher adjusted risk ratios were found among women of the Other South Asians group (145, 95% confidence interval 120–175). Survival (adjusted hazard ratio) following lower respiratory tract infection for Pakistani men (54, 95% confidence interval 39–74) and women (31, 95% confidence interval 18–53) was better than the reference population. Conclusions: Substantial differences in the rates of lower respiratory tract infections amongst different ethnic groups in Scotland were found. Pakistani men and women had particularly high rates of lower respiratory tract infection
hospitalisation. The reasons behind the high rates of lower respiratory tract infection in the Pakistani community are now required.
Keywords Respiratory tract infections, secondary care, death, ethnic groups, incidence, hospital readmission, hospitalisation
Introduction Acute lower respiratory infections are associated with significant morbidity and mortality and are one of the leading causes of years of life lost due to premature mortality.1 Lower respiratory tract infection is an umbrella term and includes (among other diseases) influenza and pneumonia. In the UK, disease rates (in particular pneumonia and influenza) usually peak during the winter and spring months (i.e. September to February and March to May).2 Risk factors that increase the likelihood of acute lower respiratory tract infections include chronic respiratory diseases (e.g. asthma and chronic obstructive pulmonary disease), tobacco smoking and immune suppression.3 Of the lower respiratory tract infections, pneumonia causes the most deaths, with four million each year worldwide. Pneumonia is particularly prevalent in children and the elderly.4–6 For lower respiratory tract infections, reduction in modifiable risk factors, e.g. smoking and increased uptake of vaccination,7–9 represents an important approach to reduce illness and death. Variations in lower respiratory tract infections by ethnic group have been noted previously in the United States (US), New Zealand and Australia.10–15 These
! The Royal Society of Medicine 2015 Reprints and permissions: sagepub.co.uk/journalsPermissions.nav Downloaded from jrs.sagepub.com by guest on November 11, 2015
2
Journal of the Royal Society of Medicine 0(0) Figure 1. Overview of Record Linkage Process (CHI and census numbers were encrypted).
Census Database
Health Database CHI Number
Record Linkage Personal Identifiers Personal Identifiers
Census Number
CHI Number Census Number (Look-up Table)
Health Information
studies found disparities in the incidence of lower respiratory tract infections among American Indian and Alaska native children and pneumonia among Black US adults and Maori in New Zealand. Some of these studies, however, did not adjust for confounding socioeconomic factors.11,13,14 National studies are needed to explore heterogeneity within broad ethnic groups, to maintain valid surveillance of trends and inequalities in diseases such as pneumonia, to set priorities (e.g. vaccination strategies and smoking cessation) and to ensure equitable service delivery.16 Furthermore, ethnic variations in infectious respiratory diseases need periodic documentation. This is because disease patterns are altering over time as a result of changing environments and risk factor patterns (e.g. pollutants, housing, diet and smoking), particularly following migration.17 Also lower respiratory tract infection rates may relate to travel, specifically religious travel such as the Hajj in which many people congregate at the same site and outbreaks of airborne/droplet borne infections (e.g. influenza, Middle East Respiratory Syndrome and meningitis) can occur.18 Thus far, there have been few studies investigating ethnic variations in hospitalisations and deaths resulting from lower respiratory tract infections. Building on related work,19 we undertook a national observational study, adjusting for socioeconomic confounding to identify ethnic group differences for lower respiratory tract infections.
Methods We linked records from Scotland’s routinely collected hospital discharge records (SMR01) and the 2001 Census using probabilistic linkage methods (using information from multiple variables to maximise the
Ethnicity Information
chances of correctly linking disparate information on individuals) creating a retrospective cohort study of 4.65 million individuals; this represented over 90% of the Scottish population in 2001 (Figure 1). All personal identifiers including the encrypted Community Health Index identifier were removed from the data before analysis began. Over 80% linkage was achieved for each ethnic group.19
Ethnic groups The ethnic group categories were those of the 2001 Census (Box 1) as reported by the head of household or individuals. These categories include the largest minority ethnic groups in Scotland in 2001. Where possible, aggregation was minimised to avoid missing important group differences, but for outcomes (discussed below) for which there were small numbers, we combined Bangladeshis with Other South Asians and the Caribbean, African and Black Scottish or Other Black as African origin to minimise the risk of inadvertently breaching anonymity.
Demographic and socioeconomic factors In addition to age and sex, sociodemographic factors extracted from census data included: highest educational qualification of the individual (and separately the household), occupation, household tenure and quintile of Scottish Index of Multiple Deprivation.20 The Scottish Index of Multiple Deprivation is the Scottish Government’s official measure of area-based multiple deprivation. It is based on 31 indicators in six individual domains of current income, employment, housing (although not household crowding), health, education, skills
Downloaded from jrs.sagepub.com by guest on November 11, 2015
Simpson et al.
3
and training and geographic access to services and telecommunications.
data. Events were identified in all age groups from any of the diagnoses codes recorded on discharge from hospital or from the death certificates. A first admission was defined as a hospital admission without prior hospital admission for lower respiratory tract infection, time to death from first admission to death (any cause) and time for first admission to a second readmission were identified.
Outcomes of interest Hospital admissions and deaths for lower respiratory tract infection, pneumonia and influenza (after hospitalisation for any of these diseases) were collected for a nine-year period from May 2001 to April 2010 (Box 2). These data (SMR01) have been found to be reliable, with completeness and accuracy rates exceeding 90%.21 Their value for epidemiologic research has been repeatedly demonstrated.22 The time period chosen reflects the time between the 2001 census and the most recent available outcomes
Statistical analysis We calculated age-adjusted rates and risk ratios (adjusted risk ratio) per 100,000/year with a personyears-at-risk denominator, using Poisson regression models with robust variance. As differences between sexes were found in previous analyses,16,19 an a priori decision was taken to stratify our analysis by sex. The White Scottish population was used as the reference population (risk ratio ¼ 100) and all comparisons relate to the reference as stated or implied. We identified first events over a nine-year period: i.e. May 2001 to April 2010. No test for interaction was carried out because of the lack of statistical power, so observations on potential effect modification are interpreted as hypothesis generating and preliminary. We further adjusted risk ratios for country of birth (born in the UK or born outside the UK) and Scottish Index of Multiple Deprivation quintile as a proxy for socioeconomic status. We report overall age-adjusted survival to a cause-specific death following a first hospital admission for the same disease, for outcomes where there were sufficient numbers to allow analyses and satisfy disclosure control protocols. Survival analysis was carried out using Cox regression to calculate hazard ratios adjusted for age (adjusted hazard ratio) using the White Scottish population as reference. Survival to death and to readmission covered the nine-year follow-up period. Proportional hazards assumption was tested with the graphical method using log-log (-ln{-ln(survival)}) plots. Analyses were undertaken using the statistical
Box 1. Scotland 2001 Census – self-reported ethnic groups. White Scottish Other White British White Irish Other White Any mixed background Indian Pakistani Bangladeshi Other South Asian Caribbean African Black Scottish or Other Black Chinese Other ethnic group
Box 2. International Classification of Disease (ICD)-9 and ICD-10 codes for respiratory conditions of interest. ICD10 title
ICD9
ICD10
Lower respiratory infections (influenza, pneumonia, acute bronchitis and bronchiolitis, respiratory tuberculosis lung abscess and pleural empyema)
480–485, 486, 487, 514, 466.0, 519.8, 466.1, 513, 510, 010–012
J10–J18, J20–J22, A15–A16, J85–J86
Pneumonia
480–486 and 514
J12–J18
Influenza
487
J10–11
Downloaded from jrs.sagepub.com by guest on November 11, 2015
4
Journal of the Royal Society of Medicine 0(0)
analysis package SAS v9.3 (SAS Institute Inc., Cary, NC, USA).
Ethical approval Ethical approval was granted by the Scotland A Research Ethics Committee. Permissions for data linkage and access to the Community Health Index were obtained from the Privacy Advisory Committee of the NHS Information Services Division, the General Register Office for Scotland (now National Records Scotland) and the Community Health Index Advisory Group. Access to data, analyses and release of outputs followed a prespecified protocol to ensure confidentiality and to prevent disclosure of any potentially identifying data.
adjusted risk ratios (145; 95% confidence interval 120–175) were found among Other South Asian women (Table 2). Also (as with White Irish men), after adjustment, risk ratios among White Irish women became significantly lower than the reference population (from 103; 95% confidence interval 95– 112 to 92; 95% confidence interval 85–100). Survival for women was better in minority ethnic groups (although this was nonsignificant for Indian women). Compared to the reference population, readmission rates after a first hospitalisation for lower respiratory tract infection were significantly lower among other White British men and women and higher among Indian and Pakistani women (Table 3).
Pneumonia Results During this retrospective study with a nine-year follow-up, there were 38 million person-years at risk available for our analyses (Table 1). The majority of person-years at risk were for the White Scottish population. Lower respiratory tract infection accounted for 201,873 hospitalisations (a rate of 528 per 100,000 person-years at risk). Pneumonia accounted for 124,764 hospitalisations (324 per 100,000 person-years at risk) and influenza 2136 hospitalisations (6 per 100,000 person-years at risk). There were 51,533 deaths from lower respiratory tract infection (134 per 100,000 person years at risk); 39,780 of deaths were due to pneumonia (104 per 100,000 person years at risk).
Lower respiratory tract infection White Scottish men had a rate of lower respiratory tract infection of 541 per 100,000 patient-years. When compared with this reference population, the adjusted risk ratios for lower respiratory tract infection hospitalisation were significantly lower among Other White British and Chinese groups (Table 1) and significantly higher in Indian and Pakistani groups. In White Irish men, after adjustment for age, socioeconomic status and country of birth, the adjusted risk ratios attenuated from a significantly higher risk ratio to almost the reference value (i.e. from an unadjusted risk ratio: 117; 95% confidence interval 107–129 to an adjusted risk ratio of 101; 95% confidence interval 93–110). Survival in men following lower respiratory tract infection was mostly similar or better in minority ethnic groups, in particular Indian and Pakistani groups. In women, the results (i.e. in comparison with White Scottish women) were similar to those in men, although significantly higher
Among men, the adjusted risk ratios for pneumonia hospitalisation were low in Other White British and Chinese, who had the lowest adjusted risk ratio of pneumonia hospitalisations of any group (62; 95% confidence interval 45–84; Table 4). High adjusted risk ratios were found in the Any Mixed Background and Pakistani groups. As with lower respiratory tract infections, after adjustment, the risk ratios for pneumonia attenuated almost to the reference value in the White Irish men. Survival following pneumonia was mostly similar or better in men of minority ethnic groups. In women, significantly higher adjusted risk ratios were found among Other South Asian (148; 95% confidence interval 115–191) and lower adjusted risk ratios (90; 95% confidence interval 81–100) in Other White (Table 5). As in men, survival for women was similar or better in minority ethnic groups. Similar risks of readmission rates to lower respiratory tract infections, after a first hospitalisation for pneumonia were found (Table 3).
Influenza Influenza hospitalisation adjusted risk ratios were found to be higher in Pakistani men (235; 95% confidence interval 134, 412) than the reference group. This was the only significant group difference (in men and women) for influenza (Supplemental Table 1).
Discussion Differences in the risks of hospital admission were found for lower respiratory tract infections including pneumonia and influenza, in particular with higher risks for Pakistani men and substantially lower risks in Chinese populations. Risks of readmission were higher for Indian and Pakistani women.
Downloaded from jrs.sagepub.com by guest on November 11, 2015
Downloaded from jrs.sagepub.com by guest on November 11, 2015
1254
977
120
168
406
81
87
94
43
White Scottish
Other White British
White Irish
Other White
Any mixed background
Indian
Pakistani
Other South Asian
African Origin
Chinese
All other ethnic group
PY at risk
25,771
52,884
24,955
27,365
110,071
50,677
43,123
223,523
159,151
12,436,76
16,312,393
N
381.9
340.9
684.2
593.0
803.1
557.2
669.8
475.5
633.8
422.7
540.6
Poisson ratesa (for 100,000 PY)
70.6 (50.5, 98.9)
63.1 (46.2, 86.0)
126.6 (94.0, 170.4)
109.7 (90.0, 133.6)
148.5 (133.9, 164.8)
103.1 (88.1, 120.5)
123.9 (103.6, 148.2)
87.9 (77.9, 99.3)
117.2 (106.9, 128.5)
78.2 (70.0, 87.3)
100.0 (–)
RR and 95% CI
Poissonb relative risks Age adjusted
70.6 (54.6, 91.2)
65.8 (53.1, 81.5)
118.0 (97.6, 142.6)
107.9 (86.9, 134.0)
147.7 (132.6, 164.5)
112.6 (96.7, 131.0)
120.9 (100.7, 145.3)
90.3 (82.6, 98.7)
107.7 (99.4, 116.7)
84.5 (78.0, 91.4)
100.0 (–)
RR and 95% CI
Age and SIMD adjustedb
Note: Poisson rates and relative risk and hazard ratio by ethnic group for men. PY, person years; RR, relative risk; SIMD, Scottish index for multiple deprivation; COB, country of birth; HR, hazard ratio. a Poisson rates are derived from relatives risks with age adjustment using the White Scottish rate as the reference. b Calculated relative to the White Scottish population at 2001 census. c Over a nine-year period between 1st May 2001 and 30th April 2010.
88,191
5912
N
Ethnic group
All lower respiratory infection
Table 1. All lower respiratory tract infection hospitalisations and deaths for men.
74.8 (52.4, 106.8)
66.7 (51.2, 87.0)
129.8 (98.2, 171.6)
113.3 (92.6, 138.7)
154.1 (140.0, 169.5)
108.7 (95.0, 124.5)
118.3 (97.0, 144.1)
92.3 (81.4, 104.6)
107.2 (97.1, 118.2)
71.9 (64.5, 80.1)
100.0 (–)
RR and 95% CI
Age and COB adjustedb
73.6 (55.1, 98.2)
68.6 (55.8, 84.4)
120.1 (97.9, 147.2)
110.3 (90.0, 135.0)
151.8 (136.1, 169.4)
117.1 (101.3, 135.3)
116.8 (96.8, 141.0)
93.6 (85.4, 102.6)
100.9 (92.5, 110.1)
79.5 (73.2, 86.3)
100.0 (–)
RR and 95% CI
Age, SIMD and COB adjustedb
7 81.8 (39.0, 171.6)
18 88.2 (55.6, 140.1)
6 47.5 (21.3, 105.8)
9 56.6 (29.4, 108.8)
36 53.6 (38.7, 74.4)
20 64.2 (41.4, 99.5)
21 101.6, (66.2, 155.8)
283 93.9 (83.5, 105.6)
353 99.3 (89.4, 110.3)
1594 91.4 (86.9, 96.2)
22651 100.0 (–)
HR and 95% CIc
Survival to death from LRTI
Simpson et al. 5
Downloaded from jrs.sagepub.com by guest on November 11, 2015
70
87
51
African origin
Chinese
All other ethnic group
29,185
52,777
22,273
22,301
108,226
46,110
46,518
259,738
178,557
13,39,624
1,786,9735
PY at risk
403.9
338.2
654.7
788.8
805.7
556.3
566.1
412.7
552.9
436.2
534.5
75.6 (60.7, 94.0)
63.3 (51.1, 78.4)
122.5 (101.4, 148.0)
147.6 (122.1, 178.3)
150.7 (127.6, 178.1)
104.1 (94.4, 114.8)
105.9 (87.1, 128.7)
77.2 (70.1, 85.0)
103.4 (95.3, 112.3)
81.6 (75.6, 88.1)
100.0 (–)
RR and 95% CI
Poisson relative risksb Age adjusted
77.5 (61.0, 98.4)
66.6 (55.4, 79.9)
116.7 (97.1, 140.3)
144.0 (119.8, 173.0)
148.6 (130.7, 169.0)
112.1 (96.6, 130.1)
102.0 (80.9, 128.7)
80.4 (73.7, 87.6)
94.8 (87.9, 102.3)
87.3 (81.4, 93.6)
100.0 (–)
RR and 95% CI
Age and SIMD adjustedb
79.0 (58.8, 106.1)
66.2 (54.2, 80.9)
123.6 (99.0, 154.3)
149.1 (120.6, 184.3)
153.7 (128.4, 184.2)
107.4 (94.1, 122.7)
103.0 (82.0, 129.4)
80.3 (72.8, 88.6)
97.5 (88.7, 107.2)
77.3 (71.0, 84.1)
100.0 (–)
RR and 95% CI
Age and COB adjustedb
79.3 (61.8, 107.7)
68.1 (56.4, 82.3)
117.3 (96.4, 142.7)
144.7 (119.9, 174.7)
150.2 (131.8, 171.2)
113.9 (97.7, 132.9)
100.5 (81.1, 124.5)
82.0 (75.4, 89.1)
92.0 (84.8, 99.8)
84.9 (78.9, 91.4)
100.0 (–)
RR and 95% CI
Age, SIMD and COB adjustedb
Note: Poisson rates and relative risk and hazard ratio by ethnic group for women. PY, person years; RR, relative risk; SIMD, Scottish index for multiple deprivation; COB, country of birth; HR, hazard ratio; LRTI, lower respiratory tract infection. a Poisson rates are derived from Relatives Risks with age adjustment using the White Scottish rate as the reference. b Calculated relative to the White Scottish population at 2001 census. c Over a nine-year period between 1 May 2001 and 30 April 2010. d Results suppressed due to small numbers.
85
322
Pakistani
Other South Asian
122
858
Other White
Indian
1282
White Irish
116
6031
Other White British
Any mixed background
95516
White Scottish
Ethnic group
All lower respiratory event
Poisson ratesa (for 100,000 PY)
Table 2. All lower respiratory tract infection hospitalisations and deaths for women.
d
8
11
13
13
18
15
172
356
1491
24,458
All
41.5 (20.8, 83.1)
110.3 (61.1, 199.3)
98.7 (57.3, 170.0)
30.6 (17.8, 52.7)
83.7 (52.7, 132.9)
55.5 (33.5, 92.1)
74.4 (64.0, 86.4)
92.3 (83.2, 102.5)
88.0 (83.5, 92.7)
100.0 (–)
HR and deathsc 95% CI
Survival to death from LRTI
6 Journal of the Royal Society of Medicine 0(0)
Downloaded from jrs.sagepub.com by guest on November 11, 2015
239
172
18
26
74
13
7
16
9
White Irish
Other White
Any mixed background
Indian
Pakistani
Other South Asian
African Origin
Chinese
All other ethnic group
120.6 (62.8, 231.9)
79.8 (46.3, 137.4)
59.5 (28.4, 124.8)
85.2 (48.4, 150.1)
103.3 (80.9, 131.8)
87.2 (57.4, 132.6)
95.0 (58.2, 155.1)
95.4 (81.9, 111.2)
104.5 (91.6, 119.3)
82.5 (77.1, 88.3)
100.0 (–)
HR and 95% CI
Overall readmission
a
6
a
7
22
13
10
69
85
368
5809
Pneumonia
N of second event
a
91.7 (41.2, 204.3)
a
152.5 (72.6, 320.2)
103.0 (67.7, 156.6)
147.5 (85.6, 254.4)
116.7 (60.7, 224.6)
103.1 (81.0, 131.2)
102.6 (82.5, 127.6)
85.5 (76.7, 95.2)
100.0 (–)
HR and 95% CI
Overall readmission
Note: LRTI, lower respiratory tract infection; HR, hazard ratio; CI, confidence interval. a Results suppressed due to small numbers.
937
15409
White Scottish
Other White British
LRTI
Male ethnic group
N of second event
All Other Ethnic Group
Chinese
African Origin
Other South Asian
Pakistani
Indian
Any mixed background
Other White
White Irish
Other White British
White Scottish
Female ethnic group
Table 3. Risk of readmission for any diagnosis following first hospital admission for LRTI and pneumonia.
8
12
11
13
79
34
16
152
239
862
15633
LRTI
N of second event
97.6 (46.5, 204.9)
77.7 (43.0, 140.4)
103.0 (55.4, 191.5)
119.9 (68.0, 211.1)
159.6 (126.5, 201.3)
163.8 (113.1, 237.4)
83.0 (50.8, 135.6)
98.5 (83.7, 115.9)
112.0 (98.2, 127.7)
80.4 (74.9, 86.2)
100.0 (–)
HR and 95% CI
Overall readmission
a
a
a
a
a
a
158.9 (101.2, 249.6)
186.2 (100.1, 346.4)
a
94.3 (71.0, 125.3)
117.0 (93.4, 146.6)
72.2 (63.7, 81.7)
100.0 (–)
HR and 95% CI
Overall readmission
a
a
19
10
a
48
80
267
5333
Pneumonia
N of second event
Simpson et al. 7
635
Other White
Downloaded from jrs.sagepub.com by guest on November 11, 2015
40
46
23
African Origin
Chinese
All other ethnic group
25,843
53,089
25,116
27,527
110,919
51,006
43,255
224,653
160,540
12,50,532
16,420,136
PY at risk
245.9
193.6
363.1
368.1
424.8
320.4
504.3
328.2
428.2
290.0
339.5
72.4 (49.5, 106.1)
57.0 (34.3, 94.7)
107.0 (72.6, 157.5)
108.4 (83.2, 141.3)
125.1 (108.8, 143.8)
94.4 (76.5, 116.4)
148.5 (122.7, 179.9)
96.7 (82.6, 113.1)
126.1 (111.9, 142.1)
85.4 (76.0, 96.0)
100.0 (–)
RR and 95% CI
Poisson relative risksb Age adjusted
72.5 (50.8, 103.6)
59.1 (43.5, 80.3)
99.7 (77.5, 128.3)
106.5 (80.9, 140.3)
125.0 (109.1, 143.1)
102.9 (87.2, 121.4)
145.0 (115.7, 181.6)
99.2 (88.7, 111.0)
115.7 (104.5, 128.1)
92.4 (84.3, 101.3)
100.0 (–)
RR and 95% CI
Age and SIMD adjustedb
76.4 (47.9, 121.7)
60.4 (39.7, 92.0)
109.6 (72.5, 165.6)
111.8 (88.8, 140.7)
130.4 (114.1, 149.0)
99.8 (81.7, 122.1)
141.7 (116.5, 172.3)
101.4 (87.3, 117.8)
115.3 (101.6, 130.9)
78.5 (69.6, 88.6)
100.0 (–)
RR and 95% CI
Age and COB adjustedb
Note: Poisson rates and relative risk and hazard ratio by ethnic group for men. PY, person years; RR, relative risk; SIMD, Scottish index for multiple deprivation; COB, country of birth; HR, hazard ratio. a Poisson rates are derived from relatives risks with age adjustment using the White Scottish rate as the reference. b Calculated relative to the White Scottish population at 2001 census. c Over a nine-year period between 1 May 2001 and 30 April 2010. d Results suppressed due to small numbers.
43
184
Other South Asian
Pakistani
85
809
White Irish
Indian
3818
Other White British
76
55,751
White Scottish
Any mixed background
Pneumonia events
Ethnic group
Poisson ratesa (per 100,000 PY)
Table 4. Pneumonia hospitalisations and deaths for men.
75.4 (50.8, 111.9)
61.6 (45.4, 83.7)
101.3 (76.7, 133.8)
108.6 (84.2, 140.0)
128.8 (111.2, 149.2)
107.1 (90.6, 126.5)
140.1 (111.9, 175.3)
102.7 (91.7, 115.1)
108.6 (97.3, 121.3)
87.1 (79.0, 96.0)
100.0 (–)
RR and 95% CI
Age, SIMD and COB adjustedb
7
16
d
d
27
14
18
212
277
1231
17776
All
103.5 (49.4, 217.2)
88.2 (54.0, 144.1)
65.0 (44.6, 94.8)
66.1 (39.1, 111.6)
106.3 (67.0, 168.8)
89.8 (78.5, 102.9)
99.6 (88.4, 112.1)
92.0 (86.8, 97.5)
100.0 (–)
HR and deathsc 95% CI
Survival to death from pneumonia
8 Journal of the Royal Society of Medicine 0(0)
504
Other White
Downloaded from jrs.sagepub.com by guest on November 11, 2015
33
45
22
African Origin
Chinese
All other ethnic group
29,308
52,882
22,464
22,429
109,070
46,363
46,745
260,933
180,068
13,47,169
17,99,4602
PY at risk
216.5
211.7
367.8
484.9
375.9
307.1
378.0
269.5
368.1
287.1
321.8
67.3 (51.0, 88.8)
65.8 (51.7, 83.6)
114.3 (91.7, 142.5)
150.7 (109.0, 208.3)
116.8 (92.5, 147.7)
95.4 (78.6, 115.8)
117.5 (90.0, 153.4)
83.8 (74.7, 93.9)
114.4 (103.1, 126.9)
89.2 (81.3, 97.9)
100.0 (–)
RR and 95% CI
Poisson relative risksb Age adjusted
69.0 (50.7, 93.9)
69.0 (55.8, 85.2)
109.5 (83.7, 143.1)
147.3 (110.9, 195.6)
115.6 (95.8, 139.3)
102.3 (81.8, 128.0)
112.9 (85.0, 150.0)
87.0 (78.3, 96.6)
105.0 (94.9, 116.1)
95.3 (87.7, 103.7)
100.0 (–)
RR and 95% CI
Age and SIMD adjustedb
71.0 (50.0, 100.7)
69.8 (52.5, 92.9)
115.2 (86.6, 153.4)
152.1 (116.2, 199.1)
120.5 (95.4, 152.4)
99.9 (80.5, 124.0)
113.1 (84.1, 152.3)
88.1 (78.4, 98.9)
106.1 (93.9, 119.8)
83.3 (74.7, 92.7)
100.0 (–)
RR and 95% CI
Age and COB adjustedb
Note: Poisson rates and relative risk and hazard ratio by ethnic group for women. PY, person years; RR, relative risk; SIMD, Scottish index for multiple deprivation; COB, country of birth; HR, hazard ratio. a Poisson rates are derived from relatives risks with age adjustment using the White Scottish rate as the reference. b Calculated relative to the White Scottish population at 2001 census. c Over a nine-year period between 1 May 2001 and 30 April 2010. d Results suppressed due to small numbers.
44
119
Other South Asian
Pakistani
56
805
White Irish
Indian
3659
Other White British
65
57,902
White Scottish
Any mixed background
Pneumonia event
Ethnic group
Poisson ratesa (for 100,000 PY)
Table 5. Pneumonia hospitalisations and deaths for women.
71.2 (50.8, 99.9)
71.5 (56.3, 90.9)
110.0 (84.1, 143.9)
148.1 (114.5, 191.4)
117.8 (96.7, 143.6)
105.2 (82.7, 133.9)
110.3 (83.9, 145.0)
89.7 (80.6, 99.8)
100.3 (90.0, 111.7)
91.4 (83.5, 100.1)
100.0 (–)
RR and 95% CI
Age, SIMD and COB adjustedb
d
7
8
11
9
13
13
124
258
1146
18,703
All
51.0 (24.3, 107.0)
101.7 (50.9, 203.4)
128.2 (71.0, 231.5)
38.5 (20.0, 74.1)
97.4 (56.6, 167.9)
57.7 (33.5, 99.3)
75.7 (63.4, 90.3)
88.9 (78.6, 100.5)
89.9 (84.6, 95.4)
100.0 (–)
HR and deathsc 95% CI
Survival to death from pneumonia
Simpson et al. 9
10
Journal of the Royal Society of Medicine 0(0)
Comparison of findings with other published work The literature on ethnicity and lower respiratory tract infections, particularly among European populations, is sparse – however, risks were found to be higher in indigenous Aboriginal children in Australia10 and Native Americans in the US.11 Comparatively higher rates of community-acquired pneumonia have been found among Black adults in the US12 and among Maori in New Zealand.13 For influenza, indigenous peoples also have high rates of disease and poorer outcomes.14 The differences, however, in categorisation of race/ethnicity and system of healthcare in the US, Australia and New Zealand prevent meaningful comparisons with our data. In the UK, no difference in the rate of influenza between ethnic groups has been previously noted. However, a study by Nyland et al. set out to determine whether care pathway differences existed between non-White and White patients admitted with laboratory-confirmed pandemic influenza A (H1N1). They found that the non-White group was less likely to receive preadmission antibiotics.15 However, there were no significant differences in delayed admission, severity at presentation for admission or likelihood of severe outcome.
Main strengths and limitations of this work The strengths of Scottish Health and Ethnicity Linkage Study include the analysis at national level, the cohort design, the availability of ethnic group and sociodemographic data (including socioeconomic status) and the analyses of many ethnic groups.23 We used Scottish Index of Multiple Deprivation because hospitalisation/death from respiratory diseases is commonest in older populations and Scottish Index of Multiple Deprivation data from the Census were available for all ages. Unlike Scottish Index of Multiple Deprivation, other alternative measures such as educational and economic status are only collected from 18 to 74 years of age. We used all diagnostic data on the hospitalisation and on mortality records (so-called multi-cause coding) and not just first hospital diagnosis and underlying cause of death. Both approaches have merits and limitations. Since we used the data primarily to compare ethnic groups, we think our approach is appropriate, especially as it maximises the number of events. Limitations include: incomplete linkage, few events in some groups (particularly for influenza) reflected in wide confidence intervals, lack of community and primary care data (where the majority of less severe lower respiratory tract infections are likely to present) and lack of linkable risk factor data, e.g. smoking (although a feasibility study to link primary care
data to the Scottish Health and Ethnicity Linkage Study database is underway). There are other important missing data including key factors along the causal pathway of disease, e.g. social and genetic determinants, disease prevention, healthcare-seeking behaviour and healthcare access (e.g. higher risks could reflect an underestimated denominator or high levels of referral to hospital in some ethnic groups), potential differential treatment of people from different ethnic backgrounds by healthcare institutions (including related to sociocultural issues or difficulties with communication) and differences between use of alternative medicine and home care in different groups (including the care of people who are dying at home). Changing disease patterns (i.e. the 2009 H1N1 pandemic influenza outbreak) and specific co-morbidities such as diabetes, tuberculosis and immunosuppression, which may be high in specific ethnic groups and contribute to lower respiratory tract infection risk, were also missing. We did not have access to data to determine changing migration over time (although we had access to country of birth from the census); we were therefore unable to adjust for the length of residence in Scotland. This is likely to have an impact, as those living for longer periods in Scotland are likely to have markedly different health and socioeconomic status than more recent migrants who had being living in Scotland for a shorter period of time. Furthermore, lower risks could reflect unrecorded emigration whereby the numerator is underestimated and the denominator overestimated. As linkage was to the 2001 census, we were unable to link data to those born in Scotland after 2001. The number of children was therefore too small to allow a separate analysis. As causative lower respiratory tract infection and pneumonia organisms are different in adults and children (requiring different control measures), a stratified analysis by age should be a priority for future research. Bias may have been introduced as some groups with complex health needs, e.g. asylum seekers and the homeless, were under-represented in the 2001 census. Linkage rates were slightly lower among ethnic minorities than among the White Scottish population, increasing the risk of bias. However, there was no appreciable variation between groups in linkage rates. The 80% linkage target achieved for all ethnic groups was determined by the investigators to be both a realistic and yet valid and searching target.
Meaning of the study results: possible mechanisms and implications for clinicians and policy makers We have shown substantial differences between ethnic groups, with low risks in Chinese and surprisingly high risks of hospitalisation (but better rates of
Downloaded from jrs.sagepub.com by guest on November 11, 2015
Simpson et al.
11
survival) among Pakistani men. This finding of higher rates of a lower respiratory tract infection hospitalisation in Pakistani men is surprising. The reference White Scottish population, with a higher prevalence of smoking when compared with other ethnic groups, should be at a higher risk of respiratory diseases.24 Vaccination, in particular with the pneumococcal polysaccharide and conjugate vaccines, haemophilus influenzae (Hib) vaccine and inactivated and live attenuated influenza vaccine, is a key preventative measure for reducing the burden of pneumonia. Although few data on specific vaccine uptake among different ethnic groups living in the UK exist, it is known that areas where more non-White groups live have lower uptake of pneumococcal polysaccharide vaccine vaccination (areas with population 99–100% White: 7.8% vs. areas with population
