Skip to main content

Table 1 Diagnostic statistics of imputed suicidal ideation among Millennium Cohort Study participants at 2007 survey, n = 63,028

From: Multiple imputation validation study: addressing unmeasured survey data in a longitudinal design

  Self-reported RAN PHQ-BIN PHQ-ORD ALL-BIN ALL-ORD
True negative 60,475 58,032
(57,873–58,192) 		58,965
(58,815–59,114) 		58,103
(57,955–58,251) 		59,001
(58,844–59,157) 		58,137
(57,985–58,288)
True positive 2553 104 (73–134) 861 (778–944) 1611 (1530–1693) 967 (885–1049) 1689 (1606–1772)
False negative – 2450 (2352–2548) 1692 (1592–1792) 942 (877–1006) 1587 (1492–1682) 865 (803–927)
False positive – 2442 (2313–2571) 1510 (1394–1626) 2372 (2257–2487) 1473 (1348–1598) 2338 (2218–2457)
Prevalence 4.05 4.04 (3.79–4.29) 3.77 (3.44–4.08) 6.32 (6.01–6.64) 3.87 (3.55–4.20) 6.39 (6.07–6.71)
Sensitivity – 4 (3–5) 34 (31–37) 63 (61–65) 38 (35–41) 66 (64–68)
Specificity – 96 (96–96) 98 (97–98) 96 (96–96) 98 (97–98) 96 (96–96)
PPV – 4 (3–5) 36 (34–39) 40 (39–42) 40 (37–42) 42 (40–44)
NPV – 96 (96–96) 97 (97–97) 98 (98–99) 97 (97–98) 99 (98–99)
CPU time – – 2:32:48.60 2:26:04.96 70:55:18.03 71:09:58.09
  1. CPU: central processing unit; reported in hours, minutes, seconds; NPV: negative predictive value; PCL-C, PTSD Checklist−Civilian Version; PHQ, Patient Health Questionnaire; PPV: positive predictive value; PTSD, posttraumatic stress disorder
  2. Self-reported suicidal ideation was indicated if reported “several days” or more to “thoughts that you would be better off dead or hurting yourself in some way”
  3. RAN: suicidal ideation was randomly assigned in line with the prevalence observed in the sample in 2007
  4. PHQ-BIN model: treated suicidal ideation as a dichotomous variable and included the remaining 8 PHQ items in the imputation model
  5. PHQ-ORD model: treated suicidal ideation as a 4-level variable and included the remaining 8 PHQ items in the imputation model
  6. ALL-BIN model: treated suicidal ideation as a dichotomous variable and included the 8 PHQ items and previously identified factors from the literature: sex, age, race/ethnicity, marital status, education attainment, 10 individual items from the RAND physical functioning module, 17 individual items and PTSD screener from the PCL-C, smoking status, sleep duration, and 5 alcohol use items from the PHQ [27,28,29,30]
  7. ALL-ORD model: treated suicidal ideation as a 4-level variable and included the 8 PHQ items and previously identified factors from the literature: sex, age, race/ethnicity, marital status, education attainment, 10 individual items from the RAND physical functioning module, 17 individual items and PTSD screener from the PCL-C, smoking status, sleep duration, and 5 alcohol use items from the PHQ [27,28,29,30]
Back to article page

BMC Medical Research Methodology

ISSN: 1471-2288

Contact us