+ All Categories
Home > Documents > Basic epidemiological principles in psychiatry and psychiatric rating scales

Basic epidemiological principles in psychiatry and psychiatric rating scales

Date post: 31-Dec-2015
Category:
Upload: uriah-holmes
View: 48 times
Download: 2 times
Share this document with a friend
Description:
Basic epidemiological principles in psychiatry and psychiatric rating scales. Sean Lynch. Research Methodology and Epidemiology - 2. Research Methodology and Epidemiology -2. This is the second part of the afternoon module and today we will look at some aspects of rating scales and theory - PowerPoint PPT Presentation
54
Basic epidemiological principles in psychiatry and psychiatric rating scales Sean Lynch Sean Lynch Research Methodology and Research Methodology and Epidemiology - 2 Epidemiology - 2
Transcript

Basic epidemiological principles in psychiatry and psychiatric rating scales

Sean LynchSean Lynch

Research Methodology andResearch Methodology andEpidemiology - 2Epidemiology - 2

Research Methodology and Research Methodology and Epidemiology -2Epidemiology -2

This is the second part of the afternoon module and todayThis is the second part of the afternoon module and today

we will look at some aspects of rating scales and theorywe will look at some aspects of rating scales and theory

In psychiatryIn psychiatry

Research Methodology and Research Methodology and Epidemiology -2Epidemiology -2

Self-rated ScalesSelf-rated Scales• Quick, save interviewer timeQuick, save interviewer time

• Lack of biasLack of bias

• ReliableReliable

• Can be used for case detectionCan be used for case detection

• Can be used as outcome measuresCan be used as outcome measures

• Only as good as the questions they askOnly as good as the questions they ask

• Problems with phrasing, ease of reading, languageProblems with phrasing, ease of reading, language

• ““Order effects”Order effects”

• ““Social desirability effects”Social desirability effects”

• ““Central tendency”Central tendency”

Research Methodology and Research Methodology and Epidemiology -2Epidemiology -2

Self-rated ScalesSelf-rated Scales• Can be used for any diagnostic category, except Can be used for any diagnostic category, except

arguably less well for psychosis, cognitive impairmentarguably less well for psychosis, cognitive impairment

• Can be used to study core symptom dimensions in depth Can be used to study core symptom dimensions in depth e.g. sleep, energye.g. sleep, energy

Research Methodology and Research Methodology and Epidemiology -2Epidemiology -2

Interview - Rated ScalesInterview - Rated Scales• Can be used to help assess whether a mental disorder is Can be used to help assess whether a mental disorder is

present or notpresent or not

• Can be used to assess the severity of symptoms in a Can be used to assess the severity of symptoms in a mental disordermental disorder

• Can be used with a taxonomical system to make Can be used with a taxonomical system to make diagnosisdiagnosis

• Can also be used to study certain dimensions of Can also be used to study certain dimensions of symptoms in great depthsymptoms in great depth

• Can be highly structured with pre-determined questionsCan be highly structured with pre-determined questions

• Can be more flexible and allow more “open-ended” Can be more flexible and allow more “open-ended” questionsquestions

Research Methodology and Research Methodology and Epidemiology -2Epidemiology -2

Interview - Rated ScalesInterview - Rated Scales• Can be subject to observer biasCan be subject to observer bias

• Different raters might disagree on assessmentDifferent raters might disagree on assessment

• Can be subject to changes in behaviour of the same rater Can be subject to changes in behaviour of the same rater over timeover time

• Can be flexible and obtain supplementary informationCan be flexible and obtain supplementary information

• Probing questions can be used to ensure subject Probing questions can be used to ensure subject understand the questionsunderstand the questions

• Some say these scales can be sensitive to “clinical Some say these scales can be sensitive to “clinical change”change”

Research Methodology and Research Methodology and Epidemiology -2Epidemiology -2

Scales need to have an “anchor point” to distinguish Scales need to have an “anchor point” to distinguish between lack of pathology and an “extreme” to assess between lack of pathology and an “extreme” to assess severe pathologysevere pathology

Scales can be broad and have numerous intermediate Scales can be broad and have numerous intermediate points with phrases or words to help guide scoring of points with phrases or words to help guide scoring of severity or degreeseverity or degree

Scales can be narrow and dichotomousScales can be narrow and dichotomous

Research Methodology and Research Methodology and Epidemiology -2Epidemiology -2

Scales can suffer from “ceiling” and “floor” effects i.e. if Scales can suffer from “ceiling” and “floor” effects i.e. if measuring dimensions within less severe or more severe measuring dimensions within less severe or more severe examples of pathology e.g. depressed mood.examples of pathology e.g. depressed mood.

It can be hard to show improvement in an item where there It can be hard to show improvement in an item where there is not much opportunity to show change! e.g. trials in mild is not much opportunity to show change! e.g. trials in mild severity depressionseverity depression

Research Methodology and Research Methodology and Epidemiology -2Epidemiology -2

Properties of ScalesProperties of Scales

1. Case finding1. Case finding

Ability to detect “cases” (true positives)Ability to detect “cases” (true positives)

Ability to distinguish “non-cases” (true negatives)Ability to distinguish “non-cases” (true negatives)

Performance on these with minimum of misclassification Performance on these with minimum of misclassification i.e. low i.e. low false positive false positive and low and low false negativefalse negative

The fine tuning of a scale cut-off point or score can be The fine tuning of a scale cut-off point or score can be biased towards biased towards sensitivity (true positives/true positives and sensitivity (true positives/true positives and false negatives) false negatives) in other words not “missing” too many in other words not “missing” too many cases, or cases, or specificity (true negatives/true negatives and false specificity (true negatives/true negatives and false positives) positives) in other words having a lower number of in other words having a lower number of

““non-cases” incorrectly identified as casesnon-cases” incorrectly identified as cases

Research Methodology and Research Methodology and Epidemiology -2Epidemiology -2

Properties of ScalesProperties of Scales

2. Assessing severity2. Assessing severity

Can less severe and more severe cases be separated by the Can less severe and more severe cases be separated by the scale?scale?

Would similar scales or measures show agreement on Would similar scales or measures show agreement on severity?severity?

3. Assessing change3. Assessing change

Is a reduction or increase in score on the scale associated Is a reduction or increase in score on the scale associated with changes in the clinical picture?with changes in the clinical picture?

Research Methodology and Research Methodology and Epidemiology -2Epidemiology -2

ReliabilityReliability

““It does what it says on the tin most times”It does what it says on the tin most times”

Test-retestTest-retest

Inter-raterInter-rater

Internal ConsistencyInternal Consistency

Research Methodology and Research Methodology and Epidemiology -2Epidemiology -2

ValidityValidity

““It really tells you what is in the tin most times”It really tells you what is in the tin most times”

ConvergentConvergent

FaceFace

ConstructConstruct

A scale can be highly reliable but measure such a narrow A scale can be highly reliable but measure such a narrow concept it is clinically meaningless i.e. not validconcept it is clinically meaningless i.e. not valid

Research Methodology and Research Methodology and Epidemiology -2Epidemiology -2

DEPRESSION RATING SCALESDEPRESSION RATING SCALES

Interview-ratedInterview-rated

HAMILTON 17 and 21 itemHAMILTON 17 and 21 item

MONTGOMERY ASBERG 10 itemMONTGOMERY ASBERG 10 item

Self-ratedSelf-rated

HAD (Hospital Anxiety and Depression) 7 itemHAD (Hospital Anxiety and Depression) 7 item

Beck (BDI) 21 and 13 itemBeck (BDI) 21 and 13 item

Research Methodology and Research Methodology and Epidemiology -2Epidemiology -2

ANXIETY RATING SCALESANXIETY RATING SCALES

Interview-ratedInterview-rated

HAMILTON 14 itemHAMILTON 14 item

Self-ratedSelf-rated

HAD (Hospital Anxiety and Depression) 7 itemHAD (Hospital Anxiety and Depression) 7 item

Research Methodology and Research Methodology and Epidemiology -2Epidemiology -2

OBESSIONAL RATING SCALESOBESSIONAL RATING SCALES

Interview-ratedInterview-rated

Y-BOCS (Yale Brown) 10 itemY-BOCS (Yale Brown) 10 item

Self-ratedSelf-rated

OCI (Obsessive Compulsive Inventory) 42 itemOCI (Obsessive Compulsive Inventory) 42 item

Research Methodology and Research Methodology and Epidemiology -2Epidemiology -2

SCHIZOPHRENIA RATING SCALESSCHIZOPHRENIA RATING SCALES

Interview-ratedInterview-rated

PANSSPANSS

SANS SANS

SAPSSAPS

BPRSBPRS

Research Methodology and Research Methodology and Epidemiology -2Epidemiology -2

MANIA RATING SCALESMANIA RATING SCALES

Interview-ratedInterview-rated

YOUNG MANIA RATING SCALEYOUNG MANIA RATING SCALE

Research Methodology and Research Methodology and Epidemiology -2Epidemiology -2

DIAGNOSTIC INTERVIEWSDIAGNOSTIC INTERVIEWS

SCAN (Wing et al ) - ICDSCAN (Wing et al ) - ICD

SCID - DSMSCID - DSM

Research Methodology and Research Methodology and Epidemiology -2Epidemiology -2

Disease concepts for diagnosisDisease concepts for diagnosis• Competing paradigms were reductionist or more Competing paradigms were reductionist or more

pragmatic and multidimensionalpragmatic and multidimensional

• Hierachical approach (Foulds)Hierachical approach (Foulds)

• Multiaxial approach (DSM)Multiaxial approach (DSM)

• Now concepts of subsyndromal disorderNow concepts of subsyndromal disorder

Levels of psychological disturbance Levels of psychological disturbance - severity- severity

• Normal distressNormal distress

• MonosymptomaticMonosymptomatic - but recognisable - but recognisable

• SubsyndromalSubsyndromal - collection of several - collection of several symptoms which fails to meet diagnostic symptoms which fails to meet diagnostic criteriacriteria

• SyndromeSyndrome

Other qualifying criteriaOther qualifying criteria• Frequency and persistence of symptomFrequency and persistence of symptom

• Functional impairment or disabilityFunctional impairment or disability

• Symptom durationSymptom duration

Factors affecting agreement on Factors affecting agreement on diagnosisdiagnosis

1.1. Can have the same information but different Can have the same information but different disease conceptsdisease concepts

2.2. Can evaluate the same information in a Can evaluate the same information in a different waydifferent way

3.3. Can elicit different information from the same Can elicit different information from the same patientpatient

4.4. Can have changes in the clinical condition of Can have changes in the clinical condition of the patient at different timesthe patient at different times

Factors affecting agreement on Factors affecting agreement on diagnosisdiagnosis

1.1. Different raters have different levels of Different raters have different levels of knowledge and expertise e.g. differences knowledge and expertise e.g. differences between primary and secondary carebetween primary and secondary care

2.2. Our diagnoses have a degree of inbuilt Our diagnoses have a degree of inbuilt uncertaintyuncertainty

3.3. How confident are we that our diagnosis is How confident are we that our diagnosis is right?right?

4.4. How often will our colleagues agree with us?How often will our colleagues agree with us?

Agreement on diagnosis - case Agreement on diagnosis - case exampleexample

Man of 40 who presents to GPMan of 40 who presents to GP

Insomnia for ten daysInsomnia for ten days

Panic attacks for three weeksPanic attacks for three weeks

Irritability at work for two weeksIrritability at work for two weeks

Reduced libido for ten daysReduced libido for ten days

Agreement on diagnosis - case Agreement on diagnosis - case exampleexample

Man of 40 who presents to GPMan of 40 who presents to GP

Insomnia for ten daysInsomnia for ten days

Panic attacks for three weeksPanic attacks for three weeks

Irritability at work for two weeksIrritability at work for two weeks

Reduced libido for ten daysReduced libido for ten days

Hopelessness about future (week)Hopelessness about future (week)

Appetite reduced (two weeks)Appetite reduced (two weeks)

Reduced energy (two weeks)Reduced energy (two weeks)

Agreement on diagnosis - case Agreement on diagnosis - case exampleexample

BUT has auditory hallucinations in third person BUT has auditory hallucinations in third person for three days!for three days!

Agreement on diagnosisAgreement on diagnosis• What is a psychiatric case? “Gold standard” - What is a psychiatric case? “Gold standard” -

Kendell, ShepherdKendell, Shepherd• Inter-rater reliabilityInter-rater reliability• Intra-rater reliabilityIntra-rater reliability• Diagnostic interviews and index of definitionDiagnostic interviews and index of definition• ““Cut-offs” based on symptom severity on rating Cut-offs” based on symptom severity on rating

instrumentsinstruments• ComputerisedComputerised

Agreement on diagnosisAgreement on diagnosis• Generally improved with more severe illnessGenerally improved with more severe illness• Difficulty in milder illness levels distinguishing Difficulty in milder illness levels distinguishing

from the normal rangefrom the normal range• More difficult for certain diagnostic concepts e.g. More difficult for certain diagnostic concepts e.g.

personality disorder and new DSM Vpersonality disorder and new DSM V

Usefulness of diagnosisUsefulness of diagnosis• Categorical verus dimensional models of illnessCategorical verus dimensional models of illness• Hypertension is not “all-or-nothing” but spectrum Hypertension is not “all-or-nothing” but spectrum

form obvious disease to normal rangeform obvious disease to normal range• Rose “not important if he has it, the question is Rose “not important if he has it, the question is

how much of it he has” how much of it he has” • Bentall - distribution of psychotic symptomsBentall - distribution of psychotic symptoms

PrevalencePrevalence• The number of defined cases of disease in an The number of defined cases of disease in an

areaarea• Includes older and more recently diagnosed Includes older and more recently diagnosed

casescases• Can be influenced by the chronicity of illness Can be influenced by the chronicity of illness

more than the incidence of illnessmore than the incidence of illness• Cases can develop and remit (or die) and Cases can develop and remit (or die) and

influence prevalenceinfluence prevalence

PrevalencePrevalence• Point prevalencePoint prevalence• Period prevalencePeriod prevalence• ““Life-time rates”Life-time rates”

Prevalence - Prevalence - interpretinginterpreting changeschanges • Can be due to true changes in incidenceCan be due to true changes in incidence• Can be due to changes in effectiveness of Can be due to changes in effectiveness of

interventionsinterventions• Can be due to changes in nature or course of Can be due to changes in nature or course of

illnessillness• Can be due to changes in detection rateCan be due to changes in detection rate

Incidence Incidence • The number of new cases of illness in an area The number of new cases of illness in an area

over a period of timeover a period of time• Changes more likely to reflect influences on Changes more likely to reflect influences on

causation or associated riskcausation or associated risk• Does not in itself give information on total Does not in itself give information on total

number of cases in communitynumber of cases in community• Changes can be due to changes in detection Changes can be due to changes in detection

raterate• Problem of including relapses inadvertentlyProblem of including relapses inadvertently

Incidence Incidence • Diseases with low incidence can become Diseases with low incidence can become

prevalent in community if chronic illnessesprevalent in community if chronic illnesses• Prevalence can change without change in Prevalence can change without change in

incidence necessarily e.g. change in severity of incidence necessarily e.g. change in severity of illness or effectiveness of treatmentillness or effectiveness of treatment

Methods to assess incidence and Methods to assess incidence and prevalence prevalence

• A case register to document all contacts overA case register to document all contacts over

a defined perioda defined period• “ “Observatory” methodObservatory” method• Case notification methodsCase notification methods• Population based studiesPopulation based studies

Methods to assess incidence and Methods to assess incidence and prevalence prevalence

• These methods can have limitations in diseases These methods can have limitations in diseases with low incidence rates and prevalence rateswith low incidence rates and prevalence rates

• The first three methods are particularly prone to The first three methods are particularly prone to error if there are problems (or there is not full error if there are problems (or there is not full consensus) on the case definitionconsensus) on the case definition

• The detection rate of cases should also be The detection rate of cases should also be measured to assess accuracy (against best measured to assess accuracy (against best available standards)available standards)

Methods to assess incidence and Methods to assess incidence and prevalence prevalence

• Case ascertainment methodsCase ascertainment methods• Prodromal phasesProdromal phases• Changes to case definitionChanges to case definition• Need to have reliable and valid raw data to Need to have reliable and valid raw data to

review estimatesreview estimates

Methods to assess incidence and Methods to assess incidence and prevalence prevalence

Other problems in psychiatry:- Other problems in psychiatry:- • ““Diagnostic overlapping”Diagnostic overlapping”• Cultural influences on case ascertainmentCultural influences on case ascertainment• Co-morbidityCo-morbidity• Diagnostic stability over time Diagnostic stability over time • Social changes (“pathoplastic”) Social changes (“pathoplastic”)

Methods to assess incidence and Methods to assess incidence and prevalence prevalence

““Head count” methods in psychiatry:- Head count” methods in psychiatry:- • Case notification depends on accuracy of Case notification depends on accuracy of

diagnostic assessment and health seeking diagnostic assessment and health seeking behaviourbehaviour

• Case registers depend on capturing all cases Case registers depend on capturing all cases and do not cope well with migration effects and and do not cope well with migration effects and changes to housing or centres of populationchanges to housing or centres of population

Methods to assess incidence and Methods to assess incidence and prevalence prevalence

Survey methods in psychiatry:- Survey methods in psychiatry:- • Need to carefully define area studiedNeed to carefully define area studied• Feasible methods of case detectionFeasible methods of case detection• Often expensive as need to cover large areas Often expensive as need to cover large areas

and numbersand numbers• Need to have accurate estimate of population Need to have accurate estimate of population

basebase

Methods to assess incidence and Methods to assess incidence and prevalence prevalence

Survey methods in psychiatry:- Survey methods in psychiatry:- • Catchment area and “house to house” methodsCatchment area and “house to house” methods• Telephone methodsTelephone methods• For less prevalent conditions and low incidence For less prevalent conditions and low incidence

conditions will need to screen a large number of conditions will need to screen a large number of “normals”“normals”

• Need a socially acceptable screening tool Need a socially acceptable screening tool

Methods to assess incidence and Methods to assess incidence and prevalence prevalence

Survey methods in psychiatry:- Survey methods in psychiatry:- • Much more problematic for more severe illnessMuch more problematic for more severe illness• Problems of selection bias e.g. “cold-spots” of Problems of selection bias e.g. “cold-spots” of

participation, wrong time of dayparticipation, wrong time of day

Methods to assess incidence and Methods to assess incidence and prevalence prevalence

Other methodsOther methods• Postal questionnairePostal questionnaire• Two-stage screeningTwo-stage screening• Representative sample e.g. random selection as Representative sample e.g. random selection as

per some marketing approachesper some marketing approaches• Quota samplesQuota samples• Convenience samplesConvenience samples• Consecutive attendances Consecutive attendances

Methods to assess incidence and Methods to assess incidence and prevalence prevalence

Error rate in study has to be defined. Error rate in study has to be defined.

Common methods:-Common methods:-• Reference to “gold standard” Reference to “gold standard” • With reference to known reliability and stability With reference to known reliability and stability

of case definitionof case definition• Random resamplingRandom resampling• Non-participation and non-completion ratesNon-participation and non-completion rates• Estimates of “double counting” or “missing” Estimates of “double counting” or “missing”

casescases

CausationCausation

Confounding variablesConfounding variables

““Latent” variablesLatent” variables

InteractionsInteractions

Protective factorsProtective factors

Causative factorsCausative factors

CausationCausation

Consider the following hypothesis:-Consider the following hypothesis:-The risk of lung cancer is 100 times higher inThe risk of lung cancer is 100 times higher in

men aged 30-35 who smoke than in non-smokers.men aged 30-35 who smoke than in non-smokers.

Smokers on average watch 50% more television thanSmokers on average watch 50% more television than

non-smokers. Smokers watching only the averagenon-smokers. Smokers watching only the average

amount of television of non-smokers could reduce theiramount of television of non-smokers could reduce their

risk of lung cancer by one third. Television might also berisk of lung cancer by one third. Television might also be

a causal factor in lung cancer.a causal factor in lung cancer.

CausationCausation

Is there an argument to support this conclusion?Is there an argument to support this conclusion?

Is the evidence convincing?Is the evidence convincing?

Can you see any flaws in this argument?Can you see any flaws in this argument?

Are there alternative methods of studying theAre there alternative methods of studying the

causal relationship between smoking, televisioncausal relationship between smoking, television

viewing and lung cancer?viewing and lung cancer?

Exposure and risksExposure and risks

Case control method is classical methodCase control method is classical method

It gives an indication of differences in the rate ofIt gives an indication of differences in the rate of

disease on exposure to a potential risk factordisease on exposure to a potential risk factor

Cross-sectional studies only give information onCross-sectional studies only give information on

associationassociation

Longitudinal studies give some more informationLongitudinal studies give some more information

on causationon causation

Exposure or “at risk” studies give the best qualityExposure or “at risk” studies give the best quality

information information

Exposure and risksExposure and risks

There are several assumptions:- There are several assumptions:- • An equal chance of exposure to a risk factor in all An equal chance of exposure to a risk factor in all

the population?the population?• Controls are “normal” Controls are “normal” • We can measure the level of risk or exposureWe can measure the level of risk or exposure• Exposure levels might vary in intensityExposure levels might vary in intensity• Duration of exposure might be relevantDuration of exposure might be relevant

Exposure and risksExposure and risks

Risk measures:-Risk measures:-

These attempt to quantify the increase in theThese attempt to quantify the increase in the

number or proportion of cases in a populationnumber or proportion of cases in a population

exposed to the risk factor, compared to those notexposed to the risk factor, compared to those not

exposed exposed • Odds ratioOdds ratio• Relative risk ratioRelative risk ratio• Attributable risk Attributable risk

Exposure and risksExposure and risks

• These are quite useful concepts but rely on These are quite useful concepts but rely on assessing one risk factor at a timeassessing one risk factor at a time

• In psychiatry multiple risk factors might be In psychiatry multiple risk factors might be expected. Sometimes it is as useful to assess expected. Sometimes it is as useful to assess interaction between factorsinteraction between factors

• We will discuss multivariate models in later We will discuss multivariate models in later sessionssessions

EBM terminologyEBM terminology

Adverse event CONTROL TREATMENTAdverse event CONTROL TREATMENT

YESYES aa bb

NONO cc dd

pc = proportion of controls with adverse eventpc = proportion of controls with adverse event

pc= b/ (b+d) pc= b/ (b+d)

pt = proportion of treatment group with adverse eventpt = proportion of treatment group with adverse event

pt = a/(a+c)pt = a/(a+c)

Relative risk of event RRe = pt/pcRelative risk of event RRe = pt/pc

Relative risk of no event or RRne =(1-pt/ 1-pc)Relative risk of no event or RRne =(1-pt/ 1-pc)

EBM terminologyEBM terminology

Odds ratio (OR) - (a x d) / (b x c)Odds ratio (OR) - (a x d) / (b x c)

Relative risk reduction RRR = (pc-pt)/ pc + 1-RReRelative risk reduction RRR = (pc-pt)/ pc + 1-RRe

Absolute risk reduction (ARR) / risk difference (RD) = pc-pt Absolute risk reduction (ARR) / risk difference (RD) = pc-pt

Number needed to treat NNTNumber needed to treat NNT

NNT (risk difference) = 1/RDNNT (risk difference) = 1/RD

NNT (relative risk of event) = 1 / (pc x RRR)NNT (relative risk of event) = 1 / (pc x RRR)

NNT (relative risk of no event) = 1 / (1-pc) x (RRne-1)NNT (relative risk of no event) = 1 / (1-pc) x (RRne-1)

NNT (odds ratio) = (1-(pc x (1-OR)) / (pc x (1-pc) * (1-OR))NNT (odds ratio) = (1-(pc x (1-OR)) / (pc x (1-pc) * (1-OR))

Other important related concepts Other important related concepts We will discuss these more fully whenWe will discuss these more fully when

discussing rating scalesdiscussing rating scales• SensitivitySensitivity• SpecificitySpecificity• Misclassification rateMisclassification rate• Predictive valuePredictive value• EfficiencyEfficiency


Recommended