1. Rebecca Bridge, Division of Epidemiology and Biostatistics
University of Illinois at Chicago, Chicago, IL
Identifying Pa@erns of HIV Testing in a Kenyan District Hospital
• In Kenya, HIV is still the leading cause of
morbidity and mortality.
• One national strategy is to identify new cases of
HIV through universal testing in healthcare
facilities.
• The aim of this study is to identify pa@erns of HIV
testing in the county district hospital in Kisumu,
Kenya where the HIV prevalence is ~19% (Kenya
Ministry of Health, 2014) and incidence is 2nd
highest in the country.
• We hypothesized that patients with the greatest
risk of infection would be more likely to be tested
despite recommended testing regardless of risk.
Highest risk groups include women 18-‐‑25, who
have the highest incidence rates, and patients with
known co-‐‑infections.
• The hypothesized highest risk groups were not
significantly more likely to be tested for HIV than
others.
• The utility of expanded testing and strategic testing
should be evaluated in the casualty department in
order to identify newly infected people to link to
care and adhere to the test and treat model.
• A limitation to this study is that it is not
generalizable as we only have information on
patients who are admi@ed.
• Another limitation is documentation in patient
charts. It is not recorded if someone was offered an
HIV test and whether or not they accepted, and
healthcare workers may not consistently document
when an HIV test is given or patient is known HIV
positive.
Conclusion
Results & Project Impact
• We conducted a retrospective chart review of
patient records who a@ended the casualty
department between 01/2014-‐‑01/2015 at Jaramogi
Oginga Odinga Teaching and Referral Hospital.
• We systematically sampled and abstracted
information from 5% of admi@ed patients 18+.
Wri@en charts are kept only for patients who are
admi@ed. After excluding those who had
documentation of previous HIV testing and those
known to have HIV our final sample size was 365.
• We coded casualty diagnoses using ICD-‐‑9 codes
and used hierarchy coding when there was more
than one diagnosis. We also recorded disposition,
date of admi@ance, home county, age, and sex.
• Using chi-‐‑square analysis we characterized
patients and used Poisson regression modeling to
produce the relative risk of being tested based on
casualty diagnosis and covariates.
Materials & Methods
Introduction
Kenya Ministry of Health. (2014). Kenya HIV County Profiles. National AIDS and STI Control
Programme. Retrieved from: h@p://www.nacc.or.ke/images/documents/KenyaCountyProfiles.pdf
Literature Cited
Janet Lin, MD, MPH, Supriya Mehta, MHS, PhD, Katherine Reifler,
Frank Ebai, Maseno University, and Jaramogi Oginga Odinga
Teaching and Referral Hospital
Acknowledgements
In 2014, 9,071 patients 18+ years were admi@ed from the casualty department. In the sample, 26% of
patients were tested for HIV. There was no significant difference in testing by gender (p-‐‑value=0.91)
and no significant difference between age groups (p-‐‑value=0.50). The RR of being tested for those
diagnosed with an infectious disease diagnosis was 1.34 (.84, 2.14).
Variable
HIV Tested,
N= 96
n (%)
HIV Not
Tested, N= 269
n (%)
Chi-‐‑
square
p-‐‑
value
Diagnosis, N=365
Other
Infectious
Genitourinary
Injury
Pregnancy
41 (27.2)
16 (36.4)
15 (40.5)
18 (20.7)
6 (13.0)
110 (72.8)
28 (63.6)
22 (59.5)
69 (79.3)
40 (87.0)
0.02
Sex, N=365
Female
Male
53 (26.37)
43 (26.22)
148 (73.6)
121 (73.8)
0.97
Age Categories, N=365
18-‐‑25
26-‐‑39
40-‐‑64
65+
21 (23.3)
24 (22.9)
26 (31.3)
25 (28.7)
69 (76.7)
81 (77.1)
57 (68.7)
62 (71.3)
0.50
County, N=360
Other
Homa Bay
Siaya
Kisumu**
18 (31.0)
10 (30.3)
34 (32.1)
34 (20.9)
40 (69.0)
23 (69.7)
72 (67.9)
129 (79.1)
0.16
Time Period, N=365
Jan, Feb, Mar
Apr, May, Jun
Jul, Aug, Sep
Oct, Nov, Dec
18 (22.0)
19 (22.9)
21 (23.6)
38 (34.2)
64 (78.0)
64 (77.1)
68 (76.4)
73 (65.8)
0.16
Table 1: Distribution of HIV Testing by
Variables
*Other diagnosis includes: circulatory, neurological, respiratory, digestive,
blood diseases, musculoskeletal, sense organs, and endocrine diagnoses
**Other county includes any county that wasn’t Homa Bay, Kisumu, or Siaya
Table 2: Gender Stratified Models-‐‑ Relative Risk of being
tested for HIV by covariates and controlling for covariates
Multivariable model includes casualty diagnosis, age category, county, and time period
*=Referent category
Male
Female
Variable
Crude
RR (95% CI)
N=164
Adjusted
RR (95% CI)
N=164
Crude
RR (95% CI)
N=201
Adjusted
RR (95% CI)
N=201
Diagnosis,
N=365
Other*
Infectious
Genitourinary
Injury
Pregnancy
Ref
1.61 (0.83, 3.11)
1.12 (0.39, 3.21)
0.96 (0.52, 1.78)
-‐‑
Ref
1.71 (0.91, 3.22)
1.19 (0.42, 3.41)
0.95 (0.51, 1.77)
-‐‑
Ref
1.13 (0.56, 2.25)
1.56 (0.91, 2.66)
0.50 (0.19, 1.31)
0.44 (0.19, 1.00)
Ref
1.06 (0.54, 2.07)
1.87 (1.08, 3.26)
0.53 (0.20, 1.39)
0.57 (0.23, 1.40)
Age Categories,
N=365
18-‐‑25
26-‐‑39
40-‐‑64
65+*
1.77 (0.75, 4.21)
1.05 (0.43, 2.57)
2.24 (1.04, 4.80)
Ref
1.59 (0.69, 3.65)
1.07 (0.45, 2.50)
2.31 (1.12, 4.82)
Ref
0.51 (0.28, 0.96)
0.70 (0.40, 1.23)
0.56 (0.28, 1.15)
Ref
0.50 (0.26, 0.95)
0.68 (0.38, 1.21)
0.57 (0.28, 1.15)
Ref
County, N=360
Other
Homa Bay
Siaya
Kisumu*
0.40 (0.09, 1.61)
1.47 (0.63, 3.41)
1.57 (0.89, 2.75)
Ref
0.42 (0.10, 1.65)
1.54 (0.65, 3.68)
1.57 (0.92, 2.70)
Ref
2.27 (1.32, 3.91)
1.42 (0.53, 3.80)
1.41 (0.76, 2.61)
Ref
2.30 (1.33, 3.95)
1.36 (0.57, 3.25)
1.39 (0.75, 2.54)
Ref
Time Period,
N=365
Jan, Feb, Mar*
Apr, May, Jun
Jul, Aug, Sep
Oct, Nov, Dec
Ref
1.61 (0.60, 4.34)
1.74 (0.67, 4.53)
2.50 (1.03, 6.08)
Ref
1.37 (0.53, 3.57)
1.61 (0.64, 4.07)
2.34 (0.99, 5.49)
Ref
0.82 (0.40, 1.68)
0.80 (0.39, 1.65)
1.19 (0.66, 2.14)
Ref
0.84 (0.42, 1.69)
0.79 (0.38, 1.61)
1.21 (0.68, 2.18)