How far does screening women for domestic (partner) - NIHR Health ...
How far does screening women for domestic (partner) - NIHR Health ...
How far does screening women for domestic (partner) - NIHR Health ...
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
DOI: 10.3310/hta13160 <strong>Health</strong> Technology Assessment 2009; Vol. 13: No. 16<br />
reported diagnostic accuracy no conclusion about<br />
construct validity can be drawn.<br />
Single question<br />
Connelly and colleagues 92 used the CTS as a<br />
comparator to test a single question incorporated<br />
into a hospital admission protocol: ‘Are you in a<br />
relationship in which you have been threatened,<br />
scared or hurt by someone? If yes, whom? [sic]’<br />
The specific clinical setting in which the protocol<br />
was administered is not clear. The CTS gave a<br />
prevalence of 18%. No diagnostic accuracy data<br />
were reported.<br />
Diagnostic accuracy<br />
The data <strong>for</strong> diagnostic accuracy are summarised<br />
in Table 5. The sensitivity of <strong>partner</strong> violence<br />
<strong>screening</strong> tools ranges from poor to good<br />
(9–100%). Tools that scored sensitivity greater<br />
than 85% include the HITS (both Spanish and<br />
English versions), STaT, WEB and OVAT. Specificity<br />
was good across all but one of the tools, ranging<br />
from 83% to 99%. The one outlier was the STaT<br />
scale, with a specificity of 37%. Index tools with<br />
a specificity greater than 85% included the AAS,<br />
HITS, single question ‘Are you safe at home?’,<br />
WAST, PVS, WEB and the BRFSS.<br />
© 2009 Queen’s Printer and Controller of HMSO. All rights reserved.<br />
The relatively small number of studies <strong>for</strong> each<br />
index tool and the heterogeneity of settings,<br />
demography of participants and comparator tools<br />
precluded pooling of the diagnostic accuracy data.<br />
Concurrent validity<br />
Five studies reported concurrent validity (Table 6);<br />
generally this was high, the exception being the<br />
PSAI.<br />
Reliability<br />
Six primary studies reported two types of reliability<br />
data (Table 7): (1) internal consistency, a measure<br />
based on the correlations between different items<br />
on the same test (coefficient alpha and Cronbach’s<br />
alpha); and (2) inter-rater reliability, the degree of<br />
agreement among scorers (Cohen’s kappa). Most<br />
studies reported good reliability of the tools that<br />
were tested.<br />
Sensitivity analyses<br />
If we exclude studies with three or more areas of<br />
bias as determined by the QUADAS appraisal tool,<br />
four studies remain, as shown in Table 8.<br />
Table 5 Ranked scores of sensitivity, specificity and summed sensitivity plus specificity <strong>for</strong> each index tool in order of decreasing<br />
predictive power<br />
Decreasing<br />
predictive power Sensitivity Specificity Overall (Sen + spec)<br />
↓ HITS<br />
(Spa)<br />
HITS (imp)<br />
STaT<br />
100<br />
96<br />
94.9<br />
AAS<br />
HITS (Eng)<br />
Single question<br />
99.2<br />
99<br />
HITS (imp)<br />
HITS (Spa)<br />
187<br />
186<br />
a HITS (Eng)<br />
WEB<br />
86<br />
85.9<br />
WAST<br />
PVS<br />
95.8<br />
95.6<br />
HITS (Eng)<br />
WEB<br />
185<br />
177<br />
b OVAT<br />
BRFSS<br />
PVS<br />
85.7<br />
72.4<br />
68<br />
WEB<br />
HITS (imp)<br />
BRFSS<br />
93.7<br />
91.1<br />
91<br />
89.7<br />
OVAT<br />
BRFSS<br />
PVS<br />
PVS<br />
168.8<br />
162.1<br />
150.3<br />
b 142.9<br />
PVSb 49.2 HITS (Spa) 86 WAST 142.6<br />
WAST 47 PVS 82.3 STaT 131.6<br />
AAS 31.7 OVAT 83.1 AAS 130.9<br />
Single questiona 8.8 STaT 36.6 Single questiona 104.6<br />
Computation of diagnostic accuracy data (Sherin et al., 1998 82 ).<br />
Spa, Spanish language (Chen et al., 2005 81 ); Eng, English language (Chen et al., 2005 81 ); imp, improvised.<br />
a Single question: ‘Do you feel safe at home?’<br />
b MacMillan et al., 2006. 85<br />
33