Скачать презентацию Addition of a help question improves the specificity Скачать презентацию Addition of a help question improves the specificity

b239d933bb38fc6271d13e332eded172.ppt

  • Количество слайдов: 34

Addition of a “help” question improves the specificity of GP diagnosis NAPCRG Quebec 2005 Addition of a “help” question improves the specificity of GP diagnosis NAPCRG Quebec 2005 Bruce Arroll, Felicity Goodyear-Smith, Ngaire Kerse, Tana Fishman, Jane Gunn* Department of General Practice & Primary Health Care School of Population Health Faculty of Medical & Health Science University of Auckland, New Zealand *University of Melbourne

Background l USA recommend screening for depression, UK does not Ann Intern Med. 2002; Background l USA recommend screening for depression, UK does not Ann Intern Med. 2002; 136: 765 -776 . Gilbody SM, House A, Shledon TA. Routinely administered questionnaires for depression and anxiety: a sytematic review. BMJ 2001; 322: 406 -9 l . 2 screening questions for depression (from original prime-MD questionnaire) – validated against CIDI “During the past month have you often been bothered by feeling down, depressed or hopeless? ” “During the past month have you often been bothered by little interest or pleasure in doing things? ”

Balancing sensitivity & specificity High sensitivity => Low sensitivity => false +ve High specificity Balancing sensitivity & specificity High sensitivity => Low sensitivity => false +ve High specificity => Low sensitivity => false -ve

Validation of 2 questions (TQ) TQ written form (patients with substance misuse excluded): Sensitivity Validation of 2 questions (TQ) TQ written form (patients with substance misuse excluded): Sensitivity 96% Specificity 57% * *Whooley et al, 1997. Case-finding instruments for depression. Two questions are as good as many. J Gen Intern Med, 12: 439 -45 TQ verbal form (Auckland): Sensitivity 96% Specificity 67% #Arroll B. Khin N. Kerse N. Screening for depression in primary care with two verbally asked questions: cross sectional study, BMJ. 327(7424): 1144 -6, 2003

Addition of help question Assess addition of help question to 2 questions (TQ) Is Addition of help question Assess addition of help question to 2 questions (TQ) Is this something with which you would like help? ‘No’, ‘Yes but not today’ or ‘Yes’ l Initially to direct the provider l Hopefullly improve sensitivity/specificity l Questionnaire known as TQWHQ

Aim l To validate the TQWHQ written for major depression Aim l To validate the TQWHQ written for major depression

Method l Consecutive sample of patients in GP l Gold standard CIDI n Composite Method l Consecutive sample of patients in GP l Gold standard CIDI n Composite international diagnostic interview l Interviewer in waiting room l North Shore practices in Auckland

Method l Written consent l Pt completed written TQWHQ + current psychotrophic drugs l Method l Written consent l Pt completed written TQWHQ + current psychotrophic drugs l Pt completed mood component CIDI interview l Pt had GP consult, showed TQWHQ to GP l GP asked any relevant questions l GP completed form - opinion whether pt depressed

Analysis l Yes response to either or both of TQ considered +ve response l Analysis l Yes response to either or both of TQ considered +ve response l Yes response to wanting help either today or not today considered +ve response l Calculated sensitivity, specificity & likelihood ratios l Designed & analysed according to STARD statement (Standards for Reporting of Diagnostic Accuracy)

Results l 1094 approached l 1025 consented (94% response rate) l 89 were on Results l 1094 approached l 1025 consented (94% response rate) l 89 were on medication - ineligible ( n=936) l 47 were CIDI +ve for major depression (5%) l 889 were CIDI -ve for major depression (95%)

Validity 2 Q alone CIDI +ve 192 False +ve 2 697 False -ve True Validity 2 Q alone CIDI +ve 192 False +ve 2 697 False -ve True -ve 47 pts major depression Both Q -ve 45 True +ve 1 or both Q +ve CIDI -ve 889 pts no major depression Sensitivity 96% Specificity 78% Miss very few; large number false +ves

Validity Help Q alone CIDI +ve 51 False +ve 12 838 False -ve True Validity Help Q alone CIDI +ve 51 False +ve 12 838 False -ve True -ve 47 pts major depression Help Q -ve 35 True +ve Help Q +ve CIDI -ve 889 pts no major depression Sensitivity 75% Specificity 94% Miss more cases but less false +ves

Validity either Q + Help Q CIDI +ve 94 False +ve 2 795 False Validity either Q + Help Q CIDI +ve 94 False +ve 2 795 False -ve True -ve 47 pts major depression Q +HQ -ve 45 True +ve Q +HQ +ve CIDI -ve 889 pts no major depression Sensitivity 96% Specificity 89%

Summary of sensitivity & specificity Sensitivity Specificity 95% CI GP diagnosis 2003 77 (58 Summary of sensitivity & specificity Sensitivity Specificity 95% CI GP diagnosis 2003 77 (58 -89 ) 86 (82 -89) GP diagnosis 2005 79 (65 -88) 94 (92 -95) TQ alone written 96 (86 -99) 78 (76 -81) HQ alone 75 (60 -85) 94 (93 -96) 2 Q or 1 Q + HQ 96 (86 -99) 89 (87 -91)

ROC Curve for Screening Questions and Diagnosis versus CIDI Ideal 100 Combination question 2 ROC Curve for Screening Questions and Diagnosis versus CIDI Ideal 100 Combination question 2 questions without help question 80 GP diagnosis -2005 GP diagnosis -2003 Sensitivity (%) Help question 60 40 20 0 0 20 40 60 1 - Specificity (%) Trade off between false-ve & false +ve rates 80 100

ROC Curve for Screening Questions & Diagnosis versus CIDI Ideal Sensitivity (%) 100 TQWHQ ROC Curve for Screening Questions & Diagnosis versus CIDI Ideal Sensitivity (%) 100 TQWHQ 80 2 questions without help question GP diagnosis -2005 GP diagnosis -2003 Help question 60 0 20 1 - Specificity (%) May not want equal tradeoff eg may prefer fewer false –ves at expenses of more false +ves 40

Likelihood ratios LR incorporates both sensitivity & specificity Direct estimate of how much test Likelihood ratios LR incorporates both sensitivity & specificity Direct estimate of how much test result changes the odds of having depression Likelihood ratio > 10 “rules in” Likelihood ratio < 0. 1 “rules out”

Likelihood ratios Answer to Help Question Help today CIDI +ve 25 CIDI -ve 27 Likelihood ratios Answer to Help Question Help today CIDI +ve 25 CIDI -ve 27 LR 17. 5 (11. 1 -27. 7) Help, but not today 10 24 7. 9 (4. 0 -15. 5) No help 12 838 0. 27 (95%CI) (0. 17 -0. 44) This means LR of a pt asking for help today having major depression is high (17. 5) Asking few more questions can confirm or refute Dx

Compare with LR of other tests Test Diagnosis LR Venography Deep vein thrombosis 47. Compare with LR of other tests Test Diagnosis LR Venography Deep vein thrombosis 47. 5 Asking for help today Major depression 17. 5 Elevation ST segment Myocardial infarct 11. 2 Asking for help not today Major depression 7. 9 D-Dimer >1092 ng/ml Deep vein thrombosis 3. 1 Not wanting help No major depression 0. 27

Help questions alone PPV= 48% PTL- = 2% PPV: proportion of patients with +ve Help questions alone PPV= 48% PTL- = 2% PPV: proportion of patients with +ve test results correctly diagnosed PTL-: proportion of patients with -ve test results incorrectly diagnosed

2 Q or 1 Q+H PPV 32% PTL- 0. 2% PPV: proportion of patients 2 Q or 1 Q+H PPV 32% PTL- 0. 2% PPV: proportion of patients with +ve test results correctly diagnosed PTL-: proportion of patients with -ve test results incorrectly diagnosed

Workload l For every 100 patients l 12 will answer yes to 2 questions Workload l For every 100 patients l 12 will answer yes to 2 questions l 3 to 1 question + help question l 5 patients with major depression detected l Miss 0. 2 patients l Without help 25 in 100 with 2 Q or 1 Q

Discussion l TQWHQ improves GP diagnosis in terms of specificity l 2 Q or Discussion l TQWHQ improves GP diagnosis in terms of specificity l 2 Q or 1 Q with a help response is the most efficient l Would only require gold standard question of 15% of patients versus 25% without help question l Gives physician permission to “help”

Recommendation GPs good at diagnosing depression in patients they see regularly (Ma. GPIe study) Recommendation GPs good at diagnosing depression in patients they see regularly (Ma. GPIe study) NZMJ 4 April 2003, Vol 116 1 -15 Use TQWHQ: l All new patients attending general practice l Patients who have not seen GP for about 2 years l Future: part of multi-item screening tool

The End The End

Discussion l 2003 study asking TQ gave 4. 3 false +ves for every true Discussion l 2003 study asking TQ gave 4. 3 false +ves for every true +ve. l 2005 study TQWHQ gave 1. 5 false +ves for every true +ve. Provides way around large numbers of false +ves in screening test

Discussion US Prevention task Force recommends l Beck Fast Scan for primary care (BFSPC) Discussion US Prevention task Force recommends l Beck Fast Scan for primary care (BFSPC) 7 questions: LR +ve / -ve = 4 / 0. 17 l Patient health question (PHQ) for depression 9 questions: LR +ve / -ve = 19. 7 / 0. 4 Our help Q: LR +ve / -ve =17. 5 / 0. 27 GP diagnosis: LR +ve / -ve = 13 / 0. 2

Discussion l ‘Help’ question derived from Case-finding & Help Assessment Tool (CHAT) l CHAT Discussion l ‘Help’ question derived from Case-finding & Help Assessment Tool (CHAT) l CHAT for use in primary health care & community settings l Help question asked for each item

Discussion Advantages of TQWHQ: l l shorter gives permission to intervene Discussion Advantages of TQWHQ: l l shorter gives permission to intervene

Previously known l High false +ve rates related to poor specificity in screening & Previously known l High false +ve rates related to poor specificity in screening & diagnostic tests l 2 screening questions good sensitivity but poor specificity for major depression l GP diagnosis with verbally asked TQ reasonable sensitivity & specificity for major depression

Conclusion – what TQWHQ adds l Response to 2 screening questions plus whether help Conclusion – what TQWHQ adds l Response to 2 screening questions plus whether help wanted today or sometime has good sensitivity & specificity for major depression l GP diagnosis with written TQWHQ similar sensiitvity but improved specificity for major depression than without help question

Acknowledgement Funding from Oakley Mental Health Foundation Study published Arroll B, Goodyear-Smith F, Kerse Acknowledgement Funding from Oakley Mental Health Foundation Study published Arroll B, Goodyear-Smith F, Kerse N, Fishman T, Gunn J (2005). Effect of the addition of a “help” question to two screening questions on specificity for diagnosis of depression in general practice: diagnostic validity study, BMJ, 15 Sep (doi: 10. 1136/bmj. 38607. 464537. 7 C)