Diarrhoea and smoking: an analysis of decades of observational data from Bangladesh

Study context and data collection

Dhaka Hospital, located in the capital city of Bangladesh, was established in 1962
by icddr,b and has since provided free medical care to all patients. The DDSS systematically
sampled 4 % of all attending patients from 1979 to 1995 and 2 % of patients since
1996 to adjust for a more than two-fold increase in patient numbers. Using a structured
questionnaire, the system collects information on infant and young children’s feeding
practices and the use of drug and fluid therapy in the home, as well as patients’
clinical, epidemiological, etiologic, and demographic characteristics. We extracted
the relevant data, which covered the period from 1993 to 2012, from the DDSS archive.
A total of 20,914 patients with diarrhoea aged 15 years and above were enrolled into
the surveillance system during this period. Of these, 20,757 were included in the
analysis and 157 were excluded because of missing data.

Assessment of smoking status

As part of DDSS, current use of cigarettes or bidis (traditional hand-rolled cigarettes)
was recorded for each enrolled patient aged 15 years and above with diarrhoea, in
addition to the number smoked per day. Although data on parental smoking behavior
was collected for children under 15 years, this was not used in the present analysis.
Current smokers (number of cigarettes or bidis smoked per day ?1) were considered
cases while non-smokers (number smoked per day?=?0) comprised the comparison group.

Laboratory methods

A fresh whole stool specimen was collected from each patient enrolled in the DDSS
and examined in icddr,b’s central laboratory in Dhaka. Each specimen was aliquoted
into three containers and submitted for routine screening for common enteric pathogens
including Vibrio cholerae, Shigella spp., Salmonella spp., Campylobacter spp., Entamoeba histolytica, Giardia lamblia22], and rotavirus 23] using standard laboratory methods.

The antimicrobial susceptibility of Shigella spp. and Vibrio cholerae to different antimicrobial agents was determined using the disk diffusion method
(CLSI 2010) employing commercial antimicrobial discs (Oxoid, Basingstoke, United Kingdom).
While we used ampicillin (10 ?g), mecillinum (25 ?g), nalidixic acid (30 ?g), trimethoprim-sulfamethoxazole
[(TMP-SXT)], (25 ?g), and ciprofloxacin (5 ?g) antibiotic discs to test the susceptibility
of Shigella, tetracycline (30 ?g), (TMP-SXT) (25 ?g), erythromycin (15 ?g), and ciprofloxacin
(5 ?g) disks were used for Vibrio cholerae24].

Data analysis

Patients’ sociodemographic and clinical characteristics, including severity of diarrhoeal
disease symptoms (mild, moderately severe, severe or very severe), abdominal pain,
stool character (watery), length of hospitalisation (24 h) and distribution of enteric
pathogens were compared between smokers and non-smokers using the chi-square test.

Diarrhoeal disease severity was scored using a 17-point numerical scale based on the
following clinical features: duration of diarrhoea, number of stools passed in last
24 h, number of occasions of vomiting in last 24 h, fever (°C), dehydration status
and treatment received (described in greater detail by Ruuska et.al.) 25]. Disease severity was then classified as mild (?6), moderately severe (7–9), severe
(10–12) or extremely severe (?13). Duration of vomiting was not used for scoring because
of incomplete data.

Wealth quintiles were estimated using principal component analysis based on household
assets such as construction materials of the house and ownership of durable goods
including a fan, radio, television, cupboard, sanitary toilet and a luxury or ordinary
cot. Age was expressed as a binary variable, with subjects categorised being 15–30
years or over 30 years respectively. The annual US$ inflation rate was used when estimating
the family income. Other binary variables were included to code for whether subjects
habitually boiled their drinking water and whether they had taken antimicrobials before
attending hospital.

Finally, we used an ordered logistic regression model (proportional odds model) to
assess the independent effects of smoking on disease severity while controlling for
age and sociodemographic factors (wealth index and level of education) and the presence
of causative pathogens (Shigella, and Vibrio cholerae) using STATA Version 12. We performed an additional analysis by including a categorical
variable for number of cigarettes smoked daily (1?=?1–9, 2?=?10–19, 3?=??20) to determine
any dose–response relationship with diarrhoeal disease severity while controlling
for the same covariates. Given that 97 % of smokers were male, female subjects were
excluded from all multivariate analyses. The strengths of association were determined
by estimating odds ratios (OR) and 95 % confidence intervals (95 % CI) with p 0.05
considered to be statistically significant.

Ethical statement

The DDSS of icddr,b is an ongoing programme of the Dhaka Hospital which has been approved
by the Research Review Committee and the Ethical Review Committee of icddr,b. Interviews
took place only after obtaining verbal consent from either patients themselves, or,
where patients were aged 15 to 19 years, from both patients and their parents or guardians,
according to hospital policy. The questionnaire recorded when consent was given. Patients
were assured of the confidentiality of all personal data collected from them and informed
about its use for research purposes and for improving patient care. The Ethical Review
Committee approved this method for obtaining consent and is satisfied that patients
participated voluntarily, that their rights were not violated, and that personal data
were handled in a confidential manner by the hospital staff.