Modifiable risk factors for typhoid intestinal perforations during a large outbreak of typhoid fever, Kampala Uganda, 2015

We defined a case of non-traumatic IP as an onset of physician-diagnosed perforation in any part of the intestine of a patient who did not have a recent trauma or injury that could explain the perforation. A physician diagnosed perforation was defined as a non-traumatic hole in any part of the wall of the gastro-intestinal tract which lines the stomach, small intestine or large bowel. A case of TIP was onset of non-traumatic IP in the terminal ileum of a physician-diagnosed typhoid patient. A case of other IP was onset of non-traumatic IP that occurred in regions other than the terminal ileum or otherwise did not meet the definition of a TIP.

Using a standardized data abstraction form, we reviewed theatre registers and case files kept at five major hospitals (described below) in Kampala City to identify all IPs, TIPs, and other IPs that occurred between January 2013 and December 2015.

Kampala is the capital city of Uganda situated in the central region of Uganda and had an estimated population of 1,516,210 based on the 2014 census [17]. The city is made up of five administrative divisions: Central, Nakawa, Makindye, Rubaga and Kawempe. We conducted the study among Kampala city residents who had been admitted in the hospitals mentioned below for TIP and patients with a diagnosis of typhoid fever without perforation. The study was conducted at five major hospitals within Kampala: Mulago National Referral Hospital, Naguru Regional Referral Hospital, Nsambya Hospital, Mengo Hospital, and Rubaga Hospital. Mulago is a public, national referral hospital which admits patients from Kampala and all over the country for free treatment. Naguru Hospital is a public regional referral hospital which admits patients from Kampala and the surrounding districts for free treatment. Nsambya, Mengo, and Rubaga hospitals are private not-for-profit hospitals that serve Kampala and surrounding districts. We selected those hospitals because patients with IPs were most likely to be referred to these high-level facilities, since the lower level health centres may not have had the capacity to handle such cases. The outbreak investigations indicated that most of the typhoid fever case-persons were of low socio-economic status [16]; therefore they are not expected to seek healthcare services in expensive private hospitals in Kampala.

We recruited all identified TIP case-patients who were alive at the time of our investigation to participate in our case-control study from 1st-23rd December 2015. Controls were residents of Kampala City line-listed during the outbreak, who were diagnosed with typhoid fever by either TUBEX-TF or blood culture confirmation, aged ≤65 years, and did not have an IP. TUBEX-TF is a rapid in vitro diagnostic test for diagnosis of acute typhoid fever [15]. For each case, we recruited three controls, randomly selected from a line list generated during the 2015 typhoid outbreak, individually matched by sex, age, and division of residence. In determining the case-to-control ratio, we estimated that approximately 50 cases could be successfully recruited and interviewed. Assuming the proportion of exposure [clinical information (pre-existing medical condition(s) such as TB, HIV, Diabetes and Cancer; duration of illness; duration of symptoms in days; duration before seeking treatment); where the first treatment was sought (i.e., self-medication, drug shop, private clinic, or hospital/health centre); number of clinics visited; and knowledge about typhoid before they contracted the disease (i.e., whether they had heard of a disease called “typhoid” before; whether they had heard of a typhoid outbreak in Kampala during the first half of 2015] in the control group was 30%, to detect an odds ratio (OR) of ≥2.5 with 95% confidence and a power of 80%, one would need approximately 150 controls, for a case-to-control ratio of 1:3. The estimated OR of ≥2.5 was based on a study conducted in Turkey, which found ORs ranging from 1.2 to 5.0 for the different risk factors identified [18].

We used a structured questionnaire to interview the case- and control-persons in person whenever they were available, or otherwise by telephone. For study participants aged <18 years, we interviewed their parents or care-givers instead.

We collected information on the following variables: socio-demographic factors (age, sex, level of education, marital status and place of residence); clinical information (pre-existing medical condition(s) such as TB, HIV, Diabetes and Cancer; duration of illness; duration of symptoms in days; duration before seeking treatment (i.e., the number of days from the onset of symptoms to the time of receiving first treatment from any health facility); where the first treatment was sought (i.e., self-medication, drug shop, private clinic, or hospital/health centre); number of clinics visited; and knowledge about typhoid before they contracted the disease (i.e., whether they had heard of a disease called “typhoid” before; whether they had heard of a typhoid outbreak in Kampala during the first half of 2015).

We evaluated the association between each individual risk factor and TIP using conditional logistic regression to account for the matched study design, and calculated the crude odds ratio (OR_crude) and its associated 95% confidence intervals (CI) for each risk factor. To adjust for confounding, we used multivariable conditional logistic regression by including co-variates in the model, and calculated the adjusted odds ratio (OR_adj) and their associated 95% CI [19]. Co-variates that were significant at the p < 0.05 level were retained in the multivariable conditional logistic regression model.

Between January 2013 and December 2014, the number of TIPs ranged between 0 and 2 per month and remained below other IPs (Fig. 1). However, the number of TIPs dramatically increased during the outbreak period (January to June) in 2015, with a peak observed in February and March. In total, 68 cases of IPs were identified during the outbreak period. The outbreak resulted in an estimated 10,230 suspected typhoid infections [16]; therefore, the estimated minimum risk of developing TIP was 6.6 (68/10,230) per 1000 suspected infections. Of the 68 TIP case-persons identified during the outbreak period, 7 died (according to patient files and surgical log books), yielding a case-fatality rate of 10%.

Between January 2013 and December 2014, cases of other IPs remained stable, ranging from 0 to 6 cases per month. As with the cases of TIP, the number of other IPs also sharply increased in January, with a peak observed in March 2015 (Fig. 1). In total, 70 cases of other IPs were identified during the outbreak period.

Of the 58 TIP case-persons whose medical records were available, 50 (86%) agreed to participate in the case-control study. The majority (86%) of TIP case-persons were males. The mean number of days from the development of symptoms to hospitalization was 11 days (range: 1–42 days). The majority of TIP case-persons (59%) were hospitalized after 7 days of illness onset. Complications after surgery occurred in 40% of TIP case-persons, with the commonest complication (70%) being wound infection (Table 1). No cases of intestinal tuberculosis, HIV, diabetes, and cancer were reported in the study population between January to June 2015.

Of the 50 TIP cases, 45 had complete data on risk factors and were used for the univariate and conditional logistic regression analysis. Cases and controls did not differ significantly in gender, mean age, education level, and the number of clinics visited before diagnoisis of TIP (Table 1). When comparing the cases and controls by individual risk factors using conditional logistic regression, risk of TIP increased with the length of time before seeking treatment: Seeking treatment ≥10 days after illness onset (OR_crude = 13, 95%CI = 3.8–43) and 4–9 days after illness onset (OR_crude = 3.0,95%CI = 1.8–5.6) were associated with significantly increased risk for TIP, compared with seeking care within 3 days of illness onset. Similarly, self-medication before seeking formal treatment (OR_crude = 2.9,95%CI = 1.4–6.2), having not heard about the typhoid outbreak in Kampala (OR_crude = 4.4,95%CI = 1.9–10.3), having got treatment before being treated at the typhoid treatment center (OR_crude = 7.0,95%CI = 1.5–31), and having not heard about typhoid fever at all (OR_crude = 2.5,95%CI = 1.0–5.9) were also significantly associated with TIP (Table 2).

When we used conditional logistic regression to control for confounding, seeking care 10+ days after onset of symptoms remained significantly associated with TIP (OR_adj = 11, 95% CI = 1.9–61), as was having received treatment before being treated at the typhoid treatment center (OR_adj = 9.0, 95%CI = 1.1–78), and being uninformed of the current typhoid outbreak (OR_adj = 5.2, 95% CI = 1.8–15) (Table 2). The associations between other risk factors and TIP became statistically non-significant in the logistic regression analysis (Table 2).